Introduction

The Latino population has been increasing in the United States since Congress passed the 1965 Immigration Act, which ended the national origins quota system and started the new wave of immigration from Latin America (Keely 1971). The US Latino population is distinctively young, with close to 65% below age 35 in 2000 and 2010 (Durand et al. 2006). This youthful profile highlights the potential contributions of Latinos to the US labor markets, culture, and other aspects of society (Tienda and Mitchell 2006). The youthfulness of this group has a less glamorous side, as well: Latinos are vulnerable to criminal victimizations that are disproportionally concentrated among young adults. For example, Latinos aged 18–34 years account for 29% of Latino population but 50% of all Latino violent victimization in the US (BJS 2015).Footnote 1 The victimization risk is particularly high for young Latino adults residing in newer settlement areas relative to their counterparts in older settlement areas. According to the 2007–2010 National Crime Victimization Survey (NCVS), young Latino adults in new settlement areas have a violent victimization rate that is almost double that of young Latino adults in traditional settlement areas (Xie and Planty 2014). Yet, the elevated risk of violent victimization for young Latino adults in newer areas of settlement is poorly understood.Footnote 2

Previously, few researchers knew about the higher risk of violent victimization for Latinos associated with newer settlement areas because many of the critical studies of Latino victimization were either limited to large Latino gateway cities such as Los Angeles, Chicago, and Miami (e.g., Block 1985; Martinez et al. 2004; Sorenson and Telles 1991), or were based on national data that did not distinguish between areas with different settlement histories (e.g., Dugan and Apel 2003; Lauritsen and Heimer 2010; Phillips 2002; Rennison 2002; Xie 2010). Other recent research has focused on crime in the total population in newly emerging immigrant areas, but was unable to separate victimization of Latinos from that of other groups (e.g., Ferraro 2016; Ramey 2013). Even the best analyses of Latinos in new settlement areas are limited to studies of homicides and police arrests, which do not capture nonfatal and unreported violence, and thus cannot address a significant portion of violent victimization in the US (e.g., Barranco 2013; Harris and Feldmeyer 2013; Painter-Davis and Harris 2016; Shihadeh and Barranco 2013; Shihadeh and Winters 2010). In short, homicide and arrest data, generated by police, cannot speak to the full range of violence against Latinos, much of which is unknown to the police, and thus cannot give a complete picture of Latino victimization in traditional and new settlement areas in the United States.

The present study enhances our understanding of Latino victimization risk by targeting this gap in the literature. Specifically, we analyze non-lethal violent victimization among young Latino adults in new and established settlement areas using new evidence from the NCVS on incidents both reported and unreported to the police. With these analyses we assess the extent of, as well as reasons for, area differences in violent victimization. Drawing on theoretical work emphasizing the protective role of traditional communities (Harris and Feldmeyer 2013; Shihadeh and Winters 2010), we evaluate the importance of individual and neighborhood factors for understanding the risks of Latino victimization across settlement areas. To do so, we combine individual victimization data with contextual data from a (restricted) geocoded version of the NCVS from the county down to the census tract level. These multilevel analyses improve our ability to distinguish the individual and contextual factors affecting risk better than earlier studies in this research area, which primarily used aggregate crime rates for neighborhoods, cities, and other geographic units.

In the following sections of the paper, we review the current research on Latino victimization by settlement areas (“Latino Victimization in Traditional and New Settlement Areas” section) and then present hypotheses of how individual- and contextual-level factors explain higher rates of violent victimization among young Latino adults in newer settlement areas compared to traditional areas (“Theoretical Background” section). We describe the data sources (“Data and Methods” section) and results of the analyses (“Results” section). Finally, we summarize the results and discuss their implications for research (“Discussion” section).

Latino Victimization in Traditional and New Settlement Areas

The Latino population resident in the United States is distinctive in its settlement patterns and these patterns have changed over time (Kandel and Cromartie 2004; Suro and Singer 2002). In 1980, 60% of Latinos lived in just five southwestern states (California, Texas, New Mexico, Arizona, and Colorado) and 22% in New York, Florida, and Illinois. From 1980 to 2010, the number of Latinos living outside of traditional states grew from 2.7 million to 13.5 million, increasingly in areas of the Southeast, Midwest, Northwest, and Northeast.

The growth of Latino populations in new settlement areas is an outcome of the combined influences of immigration, domestic migration, and fertility. For example, Lichter and Johnson (2009), using 1990–2006 data, found that nearly one-half of Latino net migration into new destinations resulted from domestic migration, whereas immigration from Mexico and other parts of the Latin America accounted for all of the net migration gains in traditional destinations. These emerging patterns of Latino settlement have multiple contributing factors. For example, the US immigration policies and laws (such as the amnesty provisions of the 1986 Immigration Control and Reform Act), the 1980s economic recession in California and Texas, the economic boom in the South and Mountain regions, and active labor recruitment by agricultural and manufacturing industries are among the factors that draw Latinos to new destinations (Massey and Capoferro 2008; Parrado and Kandel 2008). After settlement, high Latino fertility and low mortality rates promoted the growth of Latino populations in both new and traditional destinations (Johnson and Lichter 2008). As a result of these factors, there are now a substantial number of Latinos (the majority of whom are native-born) living in new Latino settlement areas.

Past research on Latinos living in traditional areas has discovered that Latinos experience lower risks of crime and victimization than would be expected on the basis of their minority status and socioeconomic background (Martinez 2002; Sampson and Bean 2006). Homicide studies found, for instance, that Latinos are significantly less likely than Blacks to be murdered (Velez 2006). Studies also have observed that contrary to frequent speculation, increases in the Latino population in the US did not predict increases in crime rates (e.g., Lee et al. 2001; Stowell and Martinez 2009; Wadsworth 2010). In addition, Feldmeyer (2009) showed that in California and New York (two traditional Latino-receiving states), the rate of Latino immigration was negatively associated with Latino-committed robbery rates, even though it showed no significant relationship to Latino homicide and violent index rates. Lee et al. (2001) further reported a negative association between immigration and Latino homicide victimization rates in El Paso, Texas, another traditional Latino-receiving state. These findings, taken together, highlight the resilience of Latino populations in the US, as has been increasingly reported in scientific publications (e.g., Lee and Martinez 2009; Sampson 2008; Velez 2009).

Nevertheless, crime patterns in traditional Latino settlement areas may not generalize to newer settlement areas. In contrast to traditional settlement areas (especially Latino gateway cities like Chicago, Miami, Los Angeles, and New York City), data for newer areas of Latino settlement are much more difficult to collect using sample surveys because of the historically low numbers of Latinos in those areas. Researchers frequently rely on targeted focus groups and non-probability samples to examine the life experiences of Latinos in those areas. Such examples include case studies of violence among Latino youths in mid-western “new gateway” cities (Enriquez et al. 2012), studies of street violence against Latinos in New Orleans, Louisiana (Fussell 2011), and studies of intimate partner violence among Latino women in the southeastern US (Denham et al. 2007; Murdaugh et al. 2004). The number of studies in this category is small, but they found that victims in their samples frequently report having limited resources and social networks to help them cope with the stresses of life in newer settlement areas.

Extending these efforts to larger and more diverse samples, studies are now increasingly using comparative data drawn from different regions of the nation to compare the differences in violence risk between new and traditional settlement areas for Latinos (see, e.g., Barranco 2013; Painter-Davis and Harris 2016; Shihadeh and Barranco 2010; Xie and Planty 2014; also see related research on settlement patterns of immigrants, Latino or otherwise, by Ferraro 2016; Harris and Feldmeyer 2013; Ramey 2013). Homicide victimization data from the Vital Statistics are an important data source for this line of research because they provide race/ethnicity-specific measures of violence across destination types (Shihadeh and Barranco 2013). Alternatively, official crime statistics from police departments may contain Latino identifiers for arrested offenders (and sometimes also for offenders and victims), even though the data necessarily limit the types of crimes one can study (and also the study’s geographic coverage) because in most states, police-based crime data do not supply this information (see discussion by Painter-Davis and Harris 2016).Footnote 3

As a third data source, the National Crime Victimization Survey (NCVS) is increasingly being used for study of Latino victimization (e.g., Dugan and Apel 2003; Hart and Rennison 2011; Lauritsen and Heimer 2010). In neighborhoods of higher Latino and immigrant concentration, there is evidence from ethnographic studies that immigrants sometimes distrust police authorities because of former experiences with crime and justice system in their home countries (Menjívar and Bejarano 2004). This factor and other factors (such as language and cultural barriers, any negative encounters with the police, fear of the repercussions of calling the police, confusion of the police with immigration enforcement agencies, and concerns about racism and discrimination) may lead to concerns about relying on the police for help after victimization, causing the police-based crime statistics to be less accurate (Davis et al. 2001; Kubrin 2014; Michelson 2003). These patterns indicate the strength of the NCVS as an alternative to police administrative records, for the data permit the analysis of crimes not reported to the police. Using NCVS data, Xie and Planty (2014), for example, were able to corroborate the homicide research of Shihadeh and Barranco (2010) by showing a higher rate of nonlethal violence for Latinos residing in new settlement areas than for those residing in traditional areas.Footnote 4

Even with these improvements in relevant data, improving our understanding of these crime patterns still faces important challenges. To date, the published comparative analyses of homicide, arrest, and the NCVS victimization rate have almost exclusively been conducted at the aggregate level with large units (states, cities, or counties). Consequently, these analyses lack individual-level information about Latinos in those areas, and the research tells us little about the conditions of neighborhoods in which they reside. These analyses of highly aggregated data limit one’s ability to disentangle contextual effects from those of the individual household and personal traits. For example, while case studies (e.g., Denham et al. 2007; Murdaugh et al. 2004) suggest that Latinos are vulnerable to victimization in new settlement areas because of individual and contextual factors (e.g., low socioeconomic status, short residence duration, language barriers, immigration status, limited access to formal and informal sources of assistance in the neighborhood, and so on), it is not clear from the existing studies how individual factors compare in importance with contextual factors in accounting for risk. Disentangling the contribution of individual and contextual factors on victimization will require larger and more representative samples with multilevel measures of individual, household, neighborhood, and broader contextual (such as city or county) characteristics. Currently, there are no data in the United States that measure all of the risk factors in a single database, but the NCVS has many of the important risk factors and the desired multilevel structure. We, therefore, employ the NCVS in the present study to further illuminate Latinos’ experiences with violent victimization. We focus on young Latino adults because prior research (Xie and Planty 2014) and our analyses in “Results” section, as elaborated below, show that this is the group most vulnerable to violent victimization in new settlement areas.

Theoretical Background

Our analysis of the NCVS focuses on three lines of research that have assessed the victimization of Latinos: First, we consider lifestyle-routine activity perspectives, specifically the argument that marriage and employment among young Latino adults serve to buffer the negative effects of economic hardship. Then, we consider two community-level hypotheses, community co-ethnic support and immigrant revitalization, which ask whether living in communities with more Latinos and a greater proportion of foreign-born Latinos (immigrants) protects young Latino adults from victimization. Although these hypotheses were not originally developed to explain differences between new and traditional areas in victimization rates, they offer important clues as to how traditional areas of settlement may protect against victimization of Latinos in ways that the new areas of settlement may lack. We build on these hypotheses to formulate a series of testable predictions. Our goal is to evaluate the importance of measured factors that are consistent with the predictions that we derive. We recognize that no single study can fully explain risk of violence in any particular population, yet we argue that our study illuminates some important new findings and points the way for future research.

Latino Marital and Employment Statuses

The early studies of victimization emphasize individual-level factors that influence the risk of victimization (see a review by Sampson and Lauritsen 1994). Lifestyle-routine activity theories suggest that individual-level differences in age, gender, employment, income, marital status, and social roles help shape daily routines and influence victimization through exposure and guardianship (Cohen et al. 1981; Hindelang et al. 1978). These theories imply that individual or “people” characteristics themselves (compositional effects) could be the source of differences in victimization rates between new and traditional areas, apart from “place” (contextual) effects.

Research emphasizes the marital and employment statuses as buffers against the consequences of material hardship among Latinos (Sampson et al. 2005). Indeed, young adult Latinos have a rate of marriage that is double that of Blacks of the same age, and the employment rate of young adult Latinos is 17% higher than that of young adult Blacks (Ruggles et al. 2010). To the extent that marriage and employment stabilize social and economic life and reduce risky behaviors (Sampson and Laub 1993), these factors have been cited as protective factors against violence among Latinos (Lopez et al. 2012; Martinez 2002; Peterson and Krivo 2005).

The literature suggests that new settlement areas may benefit from higher levels of employment and marriage among Latinos. The growth of Latino populations in newer settlement areas (particularly those in the South and Midwest) is linked to the increased availability of job opportunities (Kandel and Cromartie 2004).Footnote 5 This suggests that employment rates would be higher among Latinos in new settlement areas than in traditional areas because many of them are attracted to the new areas because of the availability of employment. Furthermore, Leach and Bean (2008) observed that Latino migrants to new destinations are a heterogeneous group with varied marital status, counter to the speculation that migrants are unattached single persons. This occurs partly because increased job opportunities in new destinations attract married individuals as well as single persons. Consistent with these observations, we find a higher proportion of employed and married young Latino adults in new settlement areas than in traditional areas when we examine data from both the American Community Survey and the NCVS sample, as we report in “Results” section.

Yet, these observed higher rates of Latino employment and marriage in new settlement areas present a puzzle. While previous research predicts that jobs and marriages should reduce victimization risk for young adult Latinos, the data show that the risk of violence is increased in new settlement areas as compared to traditional areas. This means that while employment and marriage influence victimization, they do not explain the victimization differences between new and traditional areas. Our examination of other individual-level sociodemographic variables reveals similarly limited explanatory power of these variables in explaining area victimization differences. These observations lead to our first hypothesis: (H1) Individual-level characteristics such as marriage and employment have limited ability to explain the pattern of higher Latino victimization rates in newer settlement areas. Rather, we expect that contextual differences between new and traditional settlement areas are more important for explaining area differences in Latino victimization.

Latino Concentration and Co-Ethnic Support

Naturally, traditional Latino-receiving areas are more likely than newer settlement areas to have a greater concentration of Latinos (also called Latino population density). This population structure, as many studies suggest, may give rise to an important protective mechanism, co-ethnic support, through at least two pathways.Footnote 6 First, the concentration of Latinos in a community may be expected to increase levels of social capital arising from common language, common customs, and common networks of social relationships (Woolcock 1998). In support of this view, both Moore and Vigil (1993) and Velez-Ibanez (1993) found that Latinos rely greatly on extended kin and co-ethnic networks for child care, recreation, financial support, and other assistance. Support from co-ethnic networks and community institutions also have been shown to assist in more pressing needs such as obtaining information and finding jobs (Waldinger and Lichter 2003). Moreover, studies found a stronger sense of social belonging (Phinney 1990) and less prejudice and discrimination among residents of co-ethnic communities (Perez et al. 2008). These findings suggest that for Latinos, higher concentrations of the same ethnic group are associated with communities characterized by greater levels of social support and better access to resources. Because of these associations, living in a neighborhood of higher Latino density could mean lower risks of victimization in young Latino adults.

Second, from an economic perspective, Portes and his colleagues proposed an enclave economy hypothesis suggesting that a critical mass of co-ethnic residents enhances economic development in the labor market and thus minority workers receive returns on human capital commensurate with the returns received by workers in mainstream labor markets (Portes and Manning 1986). Similar to this view, Spener and Bean (1999) found, in a study of sixty US metropolitan areas, that the size of the Spanish-speaking population is positively linked to the profitability of self-employment among Mexican workers; that is, Mexican self-employment has a positive effect on the earnings of their co-ethnic workers when the size of the Spanish-speaking market is relatively large. This relationship may be partly because the high co-ethnic density generates opportunities or niches for the self-employment to be more diverse and profitable (Zhou 2007). For example, the presence of large numbers of co-ethnic residents helps create a market for ethnic goods and services that are not available outside of ethnic enclaves (grocery stores, restaurants, convenience stores, and so on). The size of the ethnic population, according to these views, is important for the success of the ethnic economy as well as the development of other social structures in the community (Aldrich and Waldinger 1990; Zhou 2007).

To be sure, the spatial concentration of co-ethnics may not always have the desired beneficial effects (see review by Desmond and Kubrin 2009). The employment consequences of the ethnic economy can be exploitive rather than cooperative. For example, some studies found that workers in the Los Angeles janitorial industry (Cranford 2005) and garment industry (Bonacich 1993) reported occasional or frequent experiences with exploitation such as unpaid labor, wage disparity, and unstable employment arrangements. These studies suggest that the social and economic capital advantages of the ethnic communities can be mixed, but these findings do not deny, as Waldinger (1993:700) put it, “that ethnicity is often a powerful resource for minority group members,” which often is a useful vehicle for social advancement.

Taken together, the research on co-ethnic support suggests that higher concentrations of Latino population in neighborhoods should be associated with reduced risk of victimization among young Latino adults, above and beyond the impact of individual-level characteristics. Because newer settlement areas have lower shares of Latinos, the co-ethnic support argument may help explain the difference between new and traditional settlement areas in Latino victimization rates. Our second hypothesis therefore is: (H2) Latino population density, as a measure of neighborhood potential for co-ethnic support, is expected to be lower in new settlement areas, which contributes to the higher risk of Latino victimization in new areas.

Immigrant Revitalization

As immigrants and the native-born are both contributors of the new Latino settlement areas (Johnson and Lichter 2008), it is important to distinguish between the proportion of residents who are Latinos collectively and the proportion of Latinos who are foreign-born immigrants. Obviously “high densities of Latino population” and “high densities of foreign-born Latinos” are related concepts (Logan et al. 2002); however, the presence of large number of immigrants may affect crime through somewhat different mechanisms. Holding Latino density constant, the higher density of immigrants in an area may contribute to the reduction of victimization through “immigrant revitalization” (Feldmeyer 2009; Lee and Martinez 2009; Ramey 2013). This theory states that immigrants bring strong traditional values and an entrepreneurial spirit that stimulates the development of immigrant-owned businesses and immigrant network resources that help increase economic growth and the overall standard of living (Ottaviano and Peri 2006). In addition, the concentration of immigrants in contemporary American communities promotes the development of stronger family and social ties and community civic institutions (including immigrant-serving social service organizations, churches, schools, labor corners and worker centers), resulting in more informal social control and less crime (Lee et al. 2001; Ousey and Kubrin 2009; Sampson 2008).

To capture the effect of immigrant revitalization, we use two measures: (1) Latino density (the proportion of a neighborhood’s population that is Latino) and (2) Latino immigrant concentration (the proportion of Latinos in a neighborhood that are foreign born). Based on the immigrant revitalization theory, we hypothesize: (H3) Holding constant the proportion of population that is Latino, victimization risk among young Latino adults will be reduced in communities with higher proportions of foreign-born Latinos via the effect of immigrant revitalization.

Factors Modifying the Effect of Immigrant Revitalization

There is reason to suspect that the crime-reduction effect of immigrant revitalization may be smaller in areas that lack adequate conditions to allow the community-building effect of new immigrant populations to take hold. Newer settlement areas, especially those located outside central cities or in economically more affluent places, may not experience strong revitalizing benefits of immigration because immigration to those areas is a relatively new phenomenon (Massey 2008). They may miss out on the revitalization effect because immigrants have not settled in those areas long enough, or in large enough numbers and thus, they lack the kind of social, economic, political, and institutional supports that aid in the incorporation of immigrants in more established areas (e.g., Harris and Feldmeyer 2013; Ramey 2013; Shihadeh and Barranco 2013). Under these circumstances, the crime-reduction effect of immigrant concentration (through revitalization) may be weakened. There is some evidence in the literature that supports this position. Velez and Lyons (2012) for example reviewed evidence that traditional gateway cities are better positioned to draw on the already strong immigrant networks to facilitate immigrant integration. Velez (2009) further found in Chicago that increases in proportions of immigrants were related to decreased homicide rates but only in disadvantaged neighborhoods. Using the 1995 NCVS data, Lauritsen (2001) reported that higher immigrant concentration reduced violence risks in central cities, but not in other areas.

These pieces of evidence suggest that even though high concentrations of immigrants in an area should theoretically protect against victimization, this may not be the case in new settlement areas that are unprepared to cope with a large influx of newcomers. This issue is compounded by the fact that many new settlement areas are located in areas outside central cities and are economically more affluent (see “Results” section) and consequently may lack prior experiences or established immigrant infrastructures to aid in integrating growing immigrant populations. Combining these considerations leads to our fourth and last hypothesis: (H4) Higher proportions of foreign-born Latinos in a neighborhood should be negatively associated with victimization risk among young Latino adults through immigrant revitalization, but revitalization could be less effective in new settlement areas, areas outside central cities, and areas that are socioeconomically more affluent. These context-specific effects, if they are observed, would help explain differences in Latino victimization rates between new and traditional settlement areas.

Data and Methods

Area-Identified NCVS Data

The main source of data for our study is the NCVS, which is the largest household-based crime victimization survey in the US and has served as one of the nation’s primary measures of victimization since the early 1970s (Lynch and Addington 2007). The survey has several advantages for our purposes, including: (1) it covers a wide range of geographic areas across the nation, allowing the study of Latino victimization to go beyond traditional areas of Latino settlement; (2) it features a very large sample size and therefore contains large numbers of Latinos in both traditional and new settlement areas, making a comparative analysis possible; (3) it is conducted in Spanish when necessary, which facilitates the study of Latinos; (4) it captures offenses not reported to the police, which is more than half of the violence in the US (Langton et al. 2012), and therefore produces more accurate measures of violence and reduces potential bias in victims’ decisions to call the police; and (5) it includes information on multilevel (individual and contextual) correlates of violence and thus goes beyond police data (i.e., Uniform Crime Reports) which do not include information on many individual level correlates, neighborhood characteristics, or data on non-victims. Consequently, the NCVS allows for the estimation of the effects of both individual and contextual factors for victimization risks across new and traditional settlement areas.

We used an area-identified version of the NCVS that has geographic identifiers for the counties and census tracts in which respondents lived. Using the geographic codes, we linked to the NCVS the decennial Census data and data from the American Community Survey (ACS) to create community-level variables. The 2008–2012 NCVS data were used because beginning in 2008, the NCVS fully implemented redesigned samples based on the 2000 Census. The final sample for our analysis included 23,779 person-interviews on young adults who were self-identified as Latinos (regardless of race), aged 18–34 years, and living in new and traditional Latino settlement areas as defined in the next section.

New and Traditional Areas of Latino Settlement

Previous studies have used various geographic units (states, cities, counties, and metropolitan areas) to define areas of settlement, and the choice sometimes is dependent on the data availability of individual studies (see, e.g., Harris and Feldmeyer 2013; Ramey 2013; Shihadeh and Barranco 2010; Xie and Planty 2014).Footnote 7 For our research, counties are a desirable geographical unit for the study of Latino settlement patterns because counties offer employment, housing, education, and many other resources that are important to residents (Hanson 2005). Counties therefore are an important ecological environment that defines areas of settlement. Counties also are important because they are defined for the entire US across all land-use categories, from rural to suburban to urban, going beyond the boundaries of other geographic units (such as cities and metropolitan areas) which represent urban areas but not rural America. The US Latino population is increasingly residing outside urban centers, in small towns and rural areas (Massey 2008). We therefore used counties to define areas of settlement so that the analysis is inclusive of Latinos both in and outside of urban centers.

Specifically, we defined traditional and new Latino-receiving counties based on their 1980 Latino population and Latino growth rate between 1980 and 2010 (see a similar, although more complex, typology in Suro and Singer 2002).Footnote 8 Traditional areas are counties where the 1980 Latino population exceeded the then-national average of 6.4%. They include major Latino gateway localities (including Los Angeles, Chicago, New York City, and Miami), with the majority of these counties located in the Southwest and neighboring states where Latinos traditionally have resided in large numbers (see “Appendix 1”, map a).

New areas are counties where the Latino population was initially low (<6.4% in 1980) but later (between 1980 and 2010) showed a growth rate higher than the national average growth rate (246%).Footnote 9 To more accurately capture “new” areas, we excluded a small proportion (5%) of those counties that had a growing Latino population relatively early (prior to the 1990s) so that their Latino population became relatively large (i.e., larger than the national average of the time) in or prior to 2000. This means that the growth of Latino population in our “new” areas is, as the name implies, relatively recent, though our findings are not sensitive to this sample restriction. New areas are spread throughout all regions of the nation, with a greater number of them located in the Southeast and Midwest, as the literature on new settlement areas has shown (see “Appendix 1”, map b). In 2010, approximately a quarter of young Latino adults resided in new areas and two-thirds in traditional areas. Together, the two types of areas included more than 90% of young Latino adults in the United States.

Measures

Violent Victimization

In the NCVS, respondents are asked whether they experienced violent victimization in the past six months and, if so, how many times they were victimized. We coded the incidents by offense type (rape or sexual assault, robbery, aggravated assault, and simple assault), victim-offender relationship (stranger, intimate partner, and friend or acquaintance), and crime seriousness (whether the victim sustained injury and whether a weapon was present). Our estimation of each settlement area’s victimization rates utilized the rich information from the NCVS about the incidents, including the number of times each incident occurred during each time period. For the regression analyses, however, we used dichotomous victimization outcomes rather than victimization counts because it is rare for respondents to indicate multiple violent victimizations at each 6-month interval.

Individual-Level Marital and Employment Statuses

The marital status of respondents was classified as married, divorced or separated, or otherwise not-married. Divorced or separated individuals were used as the reference category. The employment status of respondents was represented by a dichotomous measure (1 = employed, 0 = not employed).

Neighborhood-Level Latino Population Density and Immigrant Concentration

To measure Latino population density and immigrant concentration, we used census tracts to approximate neighborhoods. Latino population density is the percentage of Latino population in a respondent’s census tract. Latino immigrant concentration is the percentage of the tract’s Latino population that is foreign born. We also used the percentage of non-Latino foreign-born population in the total tract population to gauge the extent of non-Latino immigrant concentration. These measures were obtained from the 2000 Decennial Census and the American Community Survey (5-year estimates for 2006–2010, 2007–2011, and 2008–2012) and then linearly interpolated to annual values and adjusted to 2010 geographic boundaries.Footnote 10 We used the natural logarithm of these measures to reduce the skewness of the distributions.

Neighborhood Location and Socioeconomic Advantage/Disadvantage

The key variables showing the context of violent victimization measure whether a respondent lived in a new settlement area, whether the residence was inside a central city, and a composite index of neighborhood socioeconomic disadvantage. The disadvantage index was calculated as the mean of standard scores on tract poverty rate, unemployment rate, percent female-headed households with children, median household income adjusted for inflation (sign reversed), percent households with public assistance income, and percent population 25 years and over without high school diplomas (Cronbach’s α = 0.87).

Characteristics of Latino Labor and Marriage Markets as Control Variables

Our study incorporated characteristics of local labor and marriage markets apart from respondents’ own marital and employment statuses. To characterize labor and marriage markets, counties are more suitable than census tracts, as individuals search for jobs and marriage partners beyond neighborhood boundaries (Lobao and Hooks 2007). To capture the sources of Latino labor supply, we included county-level measures of Latino sub-groups including percent Mexicans, Puerto Ricans, Salvadorans, and Cubans. These variables help remedy to some degree the lack of information on Latino subgroups in the NCVS. Cubans have higher rates of entrepreneurship than the other groups (Light and Gold 2000), and these measures help control for differences among these groups in occupation, socioeconomic status, daily activities, and victimization rates caused by their differences in education, cultural values, migration histories, legal statuses, and so on (Alba et al. 1994; Stowell and Martinez 2009; Velez 2006).

The counties’ labor market potentials for Latinos were measured by percent of industries in economic-base sectors (as opposed to local-serving sectors) and percent of industries that have been traditionally regarded as key employers of Latinos including construction, manufacturing, business and other professional services, and wholesale and retail trade (Kochhar 2006). For Latinos, most of the job gains (or losses) come from the growth (or decline) of construction, business and professional services, manufacturing, and wholesale and retail trade jobs (Kochhar 2006; US Department of Labor 2012). The degree of industrial clustering in these sectors is therefore closely linked to their social well-being, and possibly their risk of victimization. Moreover, we included Latino unemployment rate as an indicator of the employability of Latinos in the county that is independent of their own employment status. The quality of the Latino marriage market was measured by Latino sex ratio (number of Latino males per 100 Latino females) and Latino divorce and separation rate (Harknett and McLanahan 2004; Oropesa et al. 1994).

Additional Control Variables

At the individual level, we controlled for previously identified risk factors of violent victimization including respondents’ age, sex, household income (1–14 scale), homeownership (owned or rented), and years of residence (Dugan and Apel 2003; Lauritsen and White 2001). At the neighborhood level, the analysis controlled for contextual risks of victimization as suggested by crime opportunity and social disorganization theories, including neighborhood population density (logged population per square mile), age structure (percent ages 18-34), family disruption (percent divorced/separated), and residential instability (percent households moved into unit less than 10 years ago, and percent vacant housing). We used the percentage of black population to control for potential racial effects on violence (Krivo and Peterson 2000). Census regional categories (South, West, Midwest, and Northeast) were added to adjust for regional differences in violence.Footnote 11

Statistical Analyses

The analyses were conducted in two stages. First, we pooled the NCVS 2008-2012 data to produce more reliable total and crime-specific estimates of the average annual rates of violent victimization for young Latino adults in new versus traditional settlement areas. Our goal in this part of the analyses was to establish the difference (or lack thereof) in victimization rates between new and traditional settlement areas. We used a measure of total violence, and then followed up with detailed analyses by violence type so that the analyses would not be dominated by any single sub-class of violence such as simple assaults. Because the NCVS provides the opportunity to define violence types other than those included in the Uniform Crime Reports (UCR), we analyzed crimes that are impossible to study using the UCR, including intimate partner violence. This enhances our ability to evaluate area differences in violence with greater precision and to potentially detect effects that may be lost in the analysis of more aggregated crime classes. In estimating victimization rates, we used weights to adjust for the NCVS complex sampling design and survey nonresponse. In the NCVS, about 3% of incidents were classified as series incidents (i.e., crimes in which six or more similar incidents occurred in a six-month period and the victim could not recall the details of each incident separately). These incidents were counted in the rates as the actual number of incidents reported by the victim, up to a maximum of 10 incidents, as recommended by Lauritsen et al. (2012). The 95% confidence intervals of the rates were calculated using the generalized variance function parameters provided by the Census Bureau (see Xie and Planty 2014).

In the second part of the analyses, we tested our hypotheses using logistic regression because the outcome variables are binary (0–1). To account for the stratified multistage cluster design of the NCVS, we used survey design variables (pseudo-strata and half sample codes) and sample weights to compute variance estimates for model parameters. We used a Taylor series linear approximation to account for the clustering of individuals within neighborhoods (Cochran 1977).Footnote 12 All county and tract variables were lagged one year to capture the context of violent victimization for the year before the violence and thus keep the temporal ordering or measurement in accord with the causal ordering implicit in our hypotheses. We tested the models for multicollinearity. All models had variance inflation factors less than 4.

Results

Area Differences in Violence Among Young Latino Adults

Table 1 reports the rates of violence by violence type in new and traditional areas. The findings show that violence rates are higher in new areas for “total violence” and for the majority of the different types of violence. Specifically, the victimization rate of total violence is significantly higher in new areas (35.4 per 1000) than in traditional areas (20.6 per 1000). New areas also show higher victimization rates for rape/sexual assault, aggravated assault, simple assault, intimate partner violence, friend/acquaintance violence, violence with injury, and violence with weapon. These rates unequivocally mark new areas as higher-risk areas for young Latino adults, as was the case in studies of homicide (e.g., Shihadeh and Barranco 2013).

Table 1 Rates of violent victimization for young Latino adults by settlement area, 2008–2012
Full size table

Yet, Table 1 also reveals some notable deviations from this pattern across settlement areas. It shows that the chance of assault by a stranger is similar in new and traditional areas (10.0 vs 8.4 per 1000, p = 0.45), whereas the chance of robbery by a stranger is lower (not higher) in new settlement areas than in traditional areas (1.6 vs 3.7 per 1000, p < 0.05).Footnote 13 These findings reveal a more complex, previously unknown, pattern of violence in new settlement areas. They show that intimate partner violence and violence by friends and some other known relationships are particularly dangerous for young Latino adults in new settlement areas. To understand these patterns, we evaluate the risk of total violence and the risk of violence by victim-offender relationship in our testing of hypotheses in subsequent sections.

Testing Hypotheses: Total Violence

Summary statistics of the study variables are listed in “Appendix 2”. These statistics show large differences in individual- and contextual-level characteristics between new and traditional settlement areas. At the individual level, young Latino adults in new areas are more likely than their traditional-area counterparts to be male, married, have a job, be slightly older, and have resided in their homes for a shorter time. At the neighborhood level, as expected, Latino population density is on average significantly lower in neighborhoods in new areas than in traditional areas (20.3 vs 56.4), with both areas showing a high degree of variation, as indicated by the large standard deviations. The new and traditional areas also show a high degree of variation in their level of Latino immigrant concentration, with the mean Latino immigrant concentration being somewhat higher in new areas than in traditional areas in the study period (44.0 vs 39.2), which reflects the growing geographic diversification of Latino immigration to new areas (Massey 2008). In addition, new-area Latinos more frequently live outside central cities and in more affluent neighborhoods.

Table 2 reports the results of hypothesis testing, using total violence as the outcome variable. We rely on graphs rather than tables to present findings more effectively, so Table 2 only reports the results for the main variables, although interested readers may refer to “Appendix 3” for the results of the full model (see column “total violence”). Briefly, Table 2 contains five models. Model 1 is a baseline model with the dummy variable for “new area” showing the elevated risk of violent victimization associated with new settlement areas, controlling only for region. The next four models correspond to the four hypotheses (H1 to H4).

Table 2 Logistic regression models of total violent victimization for young Latino adults
Full size table

In Table 2, the importance of the individual-level variables is evaluated by comparing Model 2 with Model 1. The results show that consistent with H1, adding respondents’ marital and employment statuses and other individual-level factors does little to change the relationship between residence in a new settlement area and violent victimization.

Consistent with the co-ethnic support argument (H2), Model 3 shows that neighborhood Latino population density is negatively associated with violent victimization. Also, when Latino population density is held constant, the relationship between new area and violent victimization is reduced to a small value that is within the bounds of sampling error (b = .16; SE = .17). In unreported analyses we find that Latino population density is the only variable in our study that has this type of impact on the coefficient for “new area.” Clearly, Latino population density is important for understanding victimization risk. Also, note that the effect of Latino population density is essentially the same in the full model (Model 5), holding constant individual and other contextual factors.Footnote 14

Two graphs (Fig. 1a, b) illustrate the importance of Latino population density for explaining the risks of total violence by settlement areas. Figure 1a uses a hypothetical young Latino adult with mean characteristics to plot the probabilities of total violence as predicted by Model 5. It shows that in both new and traditional areas, the neighborhood Latino population density is negatively associated with victimization risk, once other attributes of persons and neighborhoods are taken into account. When Latino population density is held the same, however, young Latino adults have similar victimization risk irrespective of their settlement area. The estimated risk is slightly higher in new areas, but the differences could easily be attributed to sampling error (p = .65).

Fig. 1
figure 1

Probability of total violent victimization by settlement area and Latino population density for young Latino adults with mean characteristics. a Latino population density is negatively associated with total violence. b The role of Latino population density in explaining the difference in total violence between new and traditional areas

Full size image

In Fig. 1b, we use Model 5 described in Table 2 to create four configurations of neighborhoods. In configuration 1, we set the values of Latino population density in new and traditional areas at their respective sample means, 20 percent and 56 percent. The estimated gap in victimization between new and traditional areas in this configuration is 9 violent victimizations per 1000 residents (32.1 − 23.1 = 9), which is a substantial difference and is unlikely to be due to sampling error. In contrast, in configurations 2 and 3, values of Latino population density in new and traditional areas are set to equal values. The estimated differences in victimization between new and traditional areas in these configurations are small and statistically insignificant. In configuration 4, when values of Latino population density are set to be considerably higher in new areas than in traditional areas, the pattern of victimization is reversed: Risk of victimization is lower (not higher) in new areas than in traditional areas. Together, these results support hypothesis H2 that Latino population density, which offers opportunities for co-ethnic support, is a key variable in explaining the difference in victimization between new and traditional areas observed in the data.

Model 4 (Table 2) supports the immigrant revitalization hypothesis (H3), by showing that even after holding constant Latino population density and other variables, the concentration of Latino immigrants in a neighborhood is negatively associated with violent victimization among young Latino adults. In contrast, non-Latino immigrant concentration is not related to the victimization of young Latino adults.

Model 5 tests the interaction terms between Latino immigrant concentration and the three contextual variables (new area, central city, and neighborhood disadvantage). None of the interaction terms (including three-way interactions not reported in the table) are large and they are all within the bounds of sampling error. Thus, we found no evidence of a context-dependent effect (hypothesis H4) in the total violence models.

Violence by Victim-Offender Relationship

How do the results hold up in the analyses of violence by victim-offender relationship? Table 3 answers this question by reporting the results for intimate partner violence (panel A), friend/acquaintance violence (panel B), and stranger violence (panel C), respectively. As with Table 2 we only report the results for the main study variables, but provide the detailed results in “Appendix 3”. Table 3 confirms a major finding from the total violence models: In all three panels in Table 3, the estimated effect of “new area” changes little when individual-level variables are added. All three models support hypothesis H1 that the difference in victimization risk in new and traditional areas is not simply a function of the individual compositional effects.

Table 3 Logistic regression models of violent victimization for young Latino adults by type of violence
Full size table

In general, the results in Table 3 also are in line with the hypotheses of co-ethnic support and immigrant revitalization (H2 and H3). The estimated effects of Latino population density are negative in all three panels, although for stranger violence, the effect is small and within the bounds of sampling error (p > .3). Figure 2 displays these patterns. Note particularly the steep negative slopes of the lines for friend and acquaintance violence, and that the lines for new and traditional areas almost completely overlap at all levels of Latino population density (meaning that when holding constant Latino population density and other variables in the model, area differences in victimization become negligible). The slopes of all the lines are clearly negative although less steep for intimate partner violence and stranger violence, with the slopes for stranger violence being within the bounds of sampling error (p > .3).

Fig. 2
figure 2

Probability of violent victimization by settlement area and Latino population density for young Latino adults with mean characteristics, by type of violence. a Intimate partner violence, b friend/acquaintance violence, c stranger violence

Full size image

Table 3 also shows in Models 5 some significant interactions of Latino immigrant concentration with, respectively, central city (in panel A) and neighborhood disadvantage (in panel B). These results provide some support for the context-dependent effect of Latino immigrant concentration on victimization, though the evidence is not consistent across all three types of violence by victim-offender relationship. We illustrate the relationships using Fig. 3, which more efficiently depicts the context-specific effects, particularly for three-way interactions (coefficients omitted). Overall, Latino immigrant concentration is clearly a protective factor for young Latino adults as Fig. 3a indicates. This effect is observed for all three forms of violence, but as Fig. 3b shows, the protective role of Latino immigrant concentration for intimate partner violence is more pronounced for young Latino adults living in central cities (i.e., the two solid lines) than for those living outside of central cities (i.e., the two dotted lines). For friend/acquaintance violence, as Fig. 3c shows, in both new and traditional areas, the protective role of Latino immigrant concentration is greater for young Latino adults living in disadvantaged areas than for those living in more affluent areas. We find no significant interactions for stranger violence (Fig. 3d).

Fig. 3
figure 3

Probability of violent victimization by settlement area and Latino immigrant concentration for young Latino adults with mean characteristics, by type of violence. a Total violence, b Intimate partner violence, c Friend/acquaintance violence, d Stranger violence

Full size image

Although we hypothesize that the protective effect of Latino immigrant concentration could be weaker for new settlement areas, Fig. 3 shows very limited supporting evidence, except for the dotted line for “new area, outside central city” (Fig. 3b), which depicts a positive (not negative) relationship between Latino immigrant concentration and violence. The dotted line in Fig. 3c for “traditional area, advantaged (neighborhood)” similarly shows a positive relationship, which stands out in comparison to the more prevalent negative relationships. These results underscore the importance of context in determining the effect of Latino immigrant concentration on violence. Overall, we have strong evidence that Latino immigrant concentration reduces violence, but there is suggestive evidence that the results may vary across context.

Discussion

This study investigates an important yet understudied phenomenon: the higher risk of violent victimization among young Latino adults residing in new settlement areas compared with young Latino adults in traditional areas. New settlement areas (“Appendix 1”, map b) are now home to 12 million Latinos, 59% of whom are native born, compared with a slightly higher percentage of native-born Latinos (62%) in traditional areas (“Appendix 1”, map a). It is well established that most police-based crime data in the US lack the ability to characterize violence risk for Latinos in new and traditional areas except for limited numbers of jurisdictions (Painter-Davis and Harris 2016). These data do not capture the full extent of non-lethal violence, and are open to possible influences of law enforcement activity and may not accurately capture actual differences in crime across jurisdictions (Lynch and Addington 2007). It is necessary to develop new and innovative approaches.

Building on prior case studies and analyses of aggregate data (see "Latino Victimization in Traditional and New Settlement Areas" section), we use new evidence from the NCVS to describe more completely the patterns of Latino violent victimization across area types. By incorporating information on the type of offenses, the characteristics of individual respondents, and the conditions of their neighborhoods, we are able to better analyze differences in victimization risk between new and traditional areas of settlement. We find that, as suggested by studies of homicide (e.g., Shihadeh and Barranco 2013), young Latino adults in new areas have higher victimization rates for total violence and for the majority of the types of violence studied. Intimate partner violence and violence perpetrated by friends and acquaintances, in particular, pose unusually high risks of victimization to young Latino adults in new settlement areas. They are nearly three or more times as likely as their traditional-area counterparts to be victims of such crimes (Table 1).

New settlement areas complicate the well-documented finding that Latinos have a strong commitment to employment, marriage, family and social ties which reduces the risk of victimization (Martinez 2002; Sampson and Bean 2006). In newer settlement areas, the protective mechanisms remain undefined, because the few studies that have compared new and traditional settlement areas (see “Latino Victimization in Traditional and New Settlement Areas” section) had only aggregate data and therefore were unable to disentangle the compositional and contextual factors that elevate the risk of victimization in new settlement areas.

Our examination of individual and contextual contributions to victimization highlights the importance of neighborhood factors for explaining differences across areas in violence risks. Specifically, our findings paint a revealing picture of the socioeconomic status of the new-area young Latino adults, including their higher levels of employment, marriage, and neighborhood socioeconomic status, compared with those of young Latino adults in traditional areas. Despite these favorable characteristics, new-area Latinos are exposed to higher risks of victimization. Our results show that, net of individual and other contextual characteristics, the presence of higher concentrations of Latinos is an important neighborhood factor that protects young Latino adults and contributes to area victimization gaps. The protective role of Latino population density is stronger for violence involving non-strangers (intimate partners, friends, and acquaintances) than it is for violence involving strangers, for which the estimated effect of Latino population density is negative but small and statistically insignificant.

Our finding that Latino population density strongly reduces victimization by non-strangers has general support from a range of studies in the Latino victimization literature. Case studies by Denham et al. (2007) report striking differences in social support between Latinos who did and did not experience intimate partner violence in a few new settlement locations. Using data from rural North Carolina, Denham et al. find that Latinos who experienced no violence by intimate partners are more likely than Latinos who experienced violence to have friends or close relatives living nearby, which implies the importance of social support. Ethnographic evidence also suggests that when Latinos settle in areas that lack Latino support networks, normative (unwritten) rules of trust and reciprocity among friends and acquaintances erode because there are limited or no informal community sanctions to enforce such rules, which in turn increases the risk for interpersonal conflicts (Conley and Bohon 2009). Latino population density thus may reduce violence because it facilitates co-ethnic support. These processes operating together may explain the observed higher risk of intimate and non-stranger violence among young Latino adults in new areas compared to traditional settlement areas.

Stranger violence, in contrast, is less effectively prevented by the presence of a higher density of Latinos in neighborhoods. Perhaps it is logical that Latino population density and co-ethnic support should be more effective in preventing violence among people who know each other (intimates, acquaintances, friends, and family), because norms of reciprocity and trust can be more effective in preventing violence within interpersonal networks. These norms are likely less effective for avoiding violence by strangers.

Alternatively, or in addition, other characteristics of the local communities in traditional areas may increase the vulnerability of Latinos to stranger violence. For example, traditional settlement areas may be more likely than newer areas to have public space and public facilities for young Latino adults to increase the opportunities for social encounters with strangers. These factors may increase crime opportunities for stranger violence, which reduces the effectiveness of Latino population density for preventing violence. Another possibility is that ethnic succession and intergroup conflict may contribute to stranger violence (Heitgerd and Bursik 1987; Suttles 1968). For example, higher levels of violence between Blacks and Latinos have been reported in Los Angeles (a traditional Latino gateway city) in neighborhoods near areas that were transitioning from being predominately Black to being predominately Latino (Hipp et al. 2009). In situations like this, the increased Latino population density is positively (not negatively) associated with violence, for the areas’ population transition may create tension and violence between ethnoracial groups, particularly among strangers, which may weaken the shielding effects of Latino population density against stranger violence. These possibilities should be explored in future investigations to help increase understanding of the area differences in violence risks documented here.

Our results also highlight the importance of accounting for the areas’ differences in immigration for studying the areas’ differences in violence risks. Our finding of a negative association between Latino immigrant concentration and the violent victimization of young Latino adults supports the growing consensus that Latino immigration is a protective, not a criminogenic, factor (Lee and Martinez 2009; Ousey and Kubrin 2009; Zatz and Smith 2012). Because our data are multilevel and include violence not reported to the police, these analyses provide additional support for existing theories in this literature which have primarily used aggregate homicide or police crime statistics to infer the crime-reduction effect of immigrant revitalization.

In addition to this overall pattern of crime-reduction, we provide evidence that factors such as neighborhood economic conditions and central city location may modify the strength of the association between Latino immigrant concentration and crime. This supports the theory that the protective or revitalizing effects of immigrant concentration are more likely to materialize in places with supportive social structures and institutions. Although we recognize that our evidence is limited and at best suggestive (Fig. 3) (e.g., Lauritsen 2001; Velez and Lyons 2012), it is sufficient to warrant further investigation of these effects with more years of data and more comprehensive measures of an area’s resources and receptiveness towards immigrants. It may be that these factors are important for explaining the difference in victimization between new and traditional settlement areas.

The limitations of our study also suggest some additional avenues for future research. First, and perhaps most important, a more complete understanding of the processes involved in co-ethnic support and immigrant revitalization hypotheses will require more direct measures of the underlying mechanisms. It would be too ambitious (and too expensive) to construct a national database containing measures of neighborhood social organizational processes and economic processes for all of the geographic areas included in our study. Nevertheless, future research may develop measures of the important theoretical constructs in smaller samples or selected areas. Studies using data from the Project on Human Development in Chicago Neighborhoods (Sampson et al. 2005) and the National Longitudinal Study of Adolescent Health (Kubrin and Desmond 2015) serve as innovative examples of approaches using surveys or mixed methods to construct measures of neighborhood social capital. These methods could be tailored to research on Latinos, taking into account unique identity and socialization processes among Latinos (Massey 2008; Zúñiga and Hernandez-Leon 2005).

Second, related to the issue of identity, Latinos in the US are a heterogeneous group (Nelson and Tienda 1985). This heterogeneity reflects diversity in the histories, cultures, political and social experiences of Mexicans, Cubans, Puerto Ricans, Salvadoran, and other groups labeled as Latino in national surveys. Moreover, Latinos can also be differentiated by citizenship, immigration status, language use, year of arrival, generational status, race and other factors (see, e.g., DiPietro and Bursik 2012; Hagan and Palloni 1999; Harris and Feldmeyer 2013; Kubrin and Desmond 2015; Stowell and Martinez 2009). As research moves forward to unpack the specific dynamics underlying differences in victimization risk across settlement areas, we also will need to consider other dimensions of the diversity among Latinos.Footnote 15 While it is not possible to study Latino subgroups with the data that we use in this study, moving in this direction is clearly an essential step for research in this area.

Moreover, the NCVS traditionally has not collected data on individual immigration status (e.g., citizen, permanent resident, temporary resident, and undocumented) and because of this, it leaves open the important question of whether individual immigrant status affects victimization, an issue to which our study cannot speak. This limitation has been hard to remedy because no national data exist to identify immigrants, their victimization, and the neighborhoods in which they reside. Currently, however, the Bureau of Justice Statistics (BJS) is conducting the NCVS Instrument Redesign and Testing Project to modernize the content of the NCVS, including revising the survey to collect data on citizenship. Citizenship is not a perfect measure of immigration, but citizenship is one indicator of the extent to which immigrants are incorporated into receiving communities (Bloemraad et al. 2008). The data, when they become available, will be an important addition to the existing US crime data collections to allow a better understanding of the geographic area differences in victimization.

Last, but not least, in order to provide a more refined view of the contrast between new and traditional settlement areas, one must go beyond cross-sectional data to examine longitudinal trends in violence influenced by population movements (immigration, emigration, and internal migration) and other geographic shifts (such as natural demographic changes caused by fertility and mortality) (see, e.g., Ferraro 2016). After all, in a broader theoretical context, the study of new Latino settlement areas (including the research on Latino co-ethnic support and immigrant revitalization) is closely related to existing notions of racial transition and community dynamics, such as the study of how the invasion and succession of new demographic groups influence neighborhoods in cities like Chicago and other localities (e.g., Bursik and Webb 1982; Covington and Taylor 1989; Kubrin and Herting 2003). The fact that areas of new settlement have higher concentrations of Latino immigrants than traditional areas (see our findings in “Testing Hypotheses: Total Violence” section), for example, suggests that the long-standing ecological models of ethnic and racial assimilation and their relationship to crime may be changing. Rather than moving initially to areas of first settlement in central cities and then dispersing to other locations, immigrants may be moving immediately to these new areas. This can pose new challenges for those areas that have not developed the institutional infrastructure that facilitates assimilation or accommodation of new populations. Moreover, Latino new settlement areas have unique features (such as lower population density and lower levels of concentrated disadvantage) which may potentially make the process of neighborhood transition different from those characteristic of more traditional gateway cities like Chicago. Overall, in studying the long-term course of Latino settlement and crime, the earlier ecological theories of land use and race relations that emphasized processes of invasion and succession may prove to be very relevant and add considerably to our understanding of differences between traditional and new Latino settlement areas.

In conclusion, we have illustrated in our study how the settlement patterns of Latinos influence their risk of violent victimization, using data not available from the police administrative records in the US. We find that Latino victimization is not predominantly the product of individual traits, but rather is influenced by community characteristics and especially the presence of Latino co-ethnic support and immigrant revitalization. As such, our findings offer important new pieces of the puzzle concerning neighborhood ethnic transition and crime. Research along these lines is sure to become increasingly important, as the US Latino population continues to grow.