The Structural Context of Homicide: Accounting for Racial Differences in Process

by Lauren J. Krivo, Ruth D. Peterson
The Structural Context of Homicide: Accounting for Racial Differences in Process
Lauren J. Krivo, Ruth D. Peterson
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License



Ohio State University Ohio State University

Previous research demonstrates differences in the processes that generate black and

white rates of criminal violence. Analyses of race-specific urban homicide offending

rates for 1990 test the hypothesis that racially different effects occur because the

crime-generating process itself is conditioned by the social situations of blacks and

whites. Results show that when African Americans and whites have similar low lev-

els of concentrated disadvantage, the effects of disadvantage and homeownership

are relatively comparable.

DESPITE a recent downturn in general that the same social conditions are at the root crime rates, criminal violence is an of violent crime for all racial groups enduring concern in urban America. But this (Sampson and Wilson 1995). Accordingly,


problem is not equally serious for all com- differences in homicide rates across racial munities and groups. One of the most glar- I and ethnic groups are thought to stem from ing differences is between the homicide lev- the considerable divergence in levels of els of blacks and whites. For example, in crime-generating social conditions. 1997 the homicide offending rate for blacks Yet recent within-race analyses challenge was 8.1 times that for whites, and this ratio this assumption; they demonstrate important has fluctuated between 6 and 9 since 1976 differences between blacks and whites in the

(U.S. Department of Justice 1999). This effects of various structural predictors of large black-white difference in lethal vio- crime (Harer and Steffensmeier 1992; LaFree lence has stimulated considerable research and Drass 1996; LaFree, Drass, and O'Day directed at understanding the disparity 1992; Messner and Golden 1992; Parker and (Harer and Steffensmeier 1992; Messner and McCall 1999; Shihadeh and Ousey 1996). Golden 1992; Sampson 1987). Most studies Why is this the case? What accounts for dif- draw from structural theories that assume ferences in the processes that generate black

and white rates of criminal violence? One possibility is that racially differing effects

Direct all correspondence to Lauren J. Krivo,

occur because the crime-generating process

Department of Sociology, Ohio State University, Columbus, OH 43210 (krivo. 1 We itself is conditioned by the social situations thank Robert L. Kaufman for consultation and of these groups. Specifically, variation in lev- comments; Shaddon McKnight, John Reynolds, els of structural disadvantage may cease to and Helen D. Rizzo for their able research assis- provide meaningful distinctions among cit- tance; Barbara House for her expert computer ies once disadvantage reaches very high lev- programming; and the ASR editors and anony-

els. As a result, additional increases in struc-

mous reviewers for their helpful suggestions.

tural disadvantage may not lead to ever

This research was supported by a grant to Lauren

higher rates of violence. If true, the implica-

J. Krivo and Ruth D. Peterson from the National

tions for blacks and whites are salient. In the

Science Foundation (SBR-9320701). A previous

majority of U.S. urban areas, African Ameri-

version of this paper was presented at the annual meeting of the American Sociological Associa- cans have extremely high levels of poverty tion, San Francisco, August 1998. and other disadvantages such that there may


REVIEW,2000,VOL.65 (AuGusT:~~~-~~S) 547

be little effect on violent crime of variation in these conditions. By contrast, disadvan- tage among whites is usually much lower and thus may be within a range in which varia- tion reflects important community differ- ences strongly associated with violent out- comes. These racial differences in structural position may account for divergence in the effects of social conditions on criminal vio- lence. By implication, when the groups do not differ, we would expect the influences of crime-generating characteristics for blacks and whites to be similar.

We test these arguments by examining models of race-specific homicide. Drawing on the above logic, we explore three hypoth- eses: (1) Theoretically important structural factors have weaker effects on homicide when disadvantage is particularly wide- spread; (2) black-white differences in the ef- fects of structural conditions reflect the fact that these groups are generally observed in different portions of the disadvantage distri- bution; and by implication (3) the effects of structural factors are similar for blacks and whites only in the small number of cities in which the two groups are comparably advan- taged.


A substantial body of research investigates the structural determinants of general homicide rates, but only in the last decade have researchers begun to explore whether general structural models apply equally well to Afri- can American and white violence (Harer and Steffensmeier 1992; Krivo and Peterson 1996; LaFree and Drass 1996; LaFree et al. 1992; Messner and Golden 1992; Parker and McCall 1999; Peterson and Krivo 1993; Phillips 1997; Sampson 1987; Shihadeh and Flynn 1996; Shihadeh and Ousey 1996, 1998; Shihadeh and Steffensmeier 1994). Because of the gravity of violence for blacks, some investigators have explored crime rates for this population exclusively (Peterson and Krivo 1993; Phillips 1997; Shihadeh and Flynn 1996; Shihadeh and Steffensmeier 1994). Compared with research on violence in general, these studies show less system- atic influence of deprivation and inequality

1 on violence among blacks. For example,

1 none has demonstrated that interracial socio- economic inequality has a significant posi- tive effect on violent crime among blacks, and only Shihadeh and Flynn (1996) report that among blacks poverty and violence are associated. These studies suggest that mod- els of general criminal violence may not ap- ply equally well to all racial groups. How- ever, single-group analyses cannot tell us whether structural conditions affect racial groups in similar or dissimilar ways. Studies that have compared the determi- nants of violent crime rates among blacks and whites tend to show that major predic- tors differ for the two groups (Harer and Steffensmeier 1992; LaFree and Drass 1996; LaFree et al. 1992; Messner and Golden 1992; Shihadeh and Ousey 1996). Harer and Steffensmeier (1992) and Shihadeh and Ousey (1996) found that intraracial inequal- ity affects rates of violence among whites but not blacks, and Messner and Golden (1992) reported that absolute deprivation in- fluences white but not black killings. LaFree and his colleagues showed that greater eco- nomic well-being significantly reduces rob- bery rates for whites, but has no influence on rates for African Americans (LaFree and Drass 1996; LaFree et al. 1992). Further, educational attainment is positively associ- ated with crime rates for blacks during times of increasing black income inequality. In contrast, increased educational attainment reduces crime among whites, but only dur- ing periods of decreasing white income in- equality. ~venstudies purporting to show that the causes of violence are comparable for blacks and whites have found important racial dif- ferences in the effects of some predictors. Sampson (1987) showed that for blacks and whites, family structure has a sizable effect on juvenile and adult urban robbery rates. Yet among adults, the influence of female- headed families on robbery is three times larger for whites than blacks (a significant difference). For youth, increased per capita income significantly reduces robbery rates for whites, but not for blacks, while welfare payments affect robbery rates for blacks but not for whites. Shihadeh and Ousey (1998) demonstrated that the link between low- skilled jobs and homicide transcends racial

lines. They also found, however, that the percentage of renters in a city has a strong effect on killings by African Americans alone, while the prevalence of high school dropouts significantly increases homicide rates only for whites.

No research to date has addressed the question of why some factors are more in- fluential in producing violence for one racial group than another. We examine the issue here by exploring the possibility that such differences in effects between blacks and whites are a result of the very disparate com- munity structures in which these two races commonly reside (Sampson and Wilson 1995). More specifically, black-white differ- ences in the influence of theoretically impor- tant determinants of crime may reflect the fact that disadvantage conditions the effects of homicide predictors, and that the two populations predominantly fall within differ- ent parts of the disadvantage distribution.

With regard to the conditioning effects of disadvantage, we argue that deprivation and other community characteristics may con- tribute less to increases in violence when dis- advantage is very high than when it is low. For example, going from a 10-percent pov- erty rate to a 20-percent poverty rate may have a greater impact on the social organiza- tion of the community, and in turn on crime, than going from a 40-percent poverty rate to a 50-percent poverty rate. In the latter case, the initial level of disadvantage is so high that further increases may matter very little. Similarly, levels of other structural charac- teristics such as stability or inequality may also have weak effects because they do not appreciably differentiate communities in such high poverty contexts. In a prior study (Krivo and Peterson 1996), we reported evi- dence consistent with this argument: We found that neighborhood poverty and rela- tively few professionals are associated with increases in violent crime, but these effects level off when disadvantage is extreme.

The argument posed here is similar to that made by scholars who have explored tipping points and other nonlinear relationships be- tween social conditions and behavioral out- comes (Crane 1991; Granovetter 1978; LaFree 1999; Quercia and Galster 1999; Schelling 1971). While much of this litera- ture focuses on how outcomes take off at a rapid rate once they reach a threshold level, our argument describes a related but oppo- site process-one of decelerating effects (Berry 1991; LaFree 1999). In particular, we hypothesize that disadvantage has a substan- tial and positive effect on homicide rates when disadvantage is low to moderate, but that this association levels off under condi- tions of extreme disadvantage.

If the nonlinear relationship described holds, effects of structural conditions for blacks and whites could differ because the observed ranges of economic deprivation for African Americans and whites are highly di- vergent and largely nonoverlapping. Depri- vation levels for blacks are generally at the high end of the distribution (where crime- producing effects level off), while depriva- tion levels for whites are in the lower por- tion of the distribution (where these effects are stronger). If this logic is correct, then crime-generating characteristics should have similar influences for blacks and whites in situations where the groups' positions are similar.

Our analyses examine race-specific homi- cide rates for large U.S. cities with appre- ciable African American populations. Metro- politan Statistical Area central cities with a population of at least 100,000 and a black population of at least 5,000 in 1990 were in- cluded to ensure a sufficient number of blacks for constructing reliable race-specific measures. Although 135 central cities met our selection criteria, homicide data were not available for cities in Florida or for Chat- tanooga, Tennessee.' Outlier analyses re- sulted in the elimination of two additional cases (Oxnard, California and Worcester, Massach~setts).~

Our final sample consists of 124 central cities.

Homicide data were obtained from the Cen- ter for the Study and Prevention of Violence (1995) at the University of Colorado. These data did not include Florida cities because Florida did not report data to the Federal Bureau of Investi- gation (FBI) from 1989 through 1991. The data set we obtained also excluded Chattanooga, Ten- nessee.

We concluded that Oxnard and Worcester are

Homicide data (murder and nonnegligent manslaughter) were drawn from the FBI's Supplementary Homicide Reports (SHR). The SHR provides data on individual homi- cide incidents, including offenders' race, which permits construction of race-specific homicide offending rates. To address the problem of missing information on race, we acquired adjusted race-specific homicide of- fender data from the Center for the Study and Prevention of Violence (1995). These adjusted counts combine the known distribu- tion of offenders' race with an estimated ra- cial distribution of offenders when race is unknown (Williams and Flewelling 1987, 1988; also see Riedel 1990).

Race-specific homicides rates are three- year averages (1989-1991) per 100,000 black (or white) population. Average rates are calculated to minimize the impact of ran- dom fluctuations in homicides from year to year. Because of skewness, the homicide rates are transformed l~garithmicall~.~

Indicators of factors identified in several major perspectives on crime are considered as independent variables: concentrated dis- advantage, community stability, racial resi- dential segregation, and interracial socio- economic inequality. Wilson (1987, 1996) and Sampson and Wilson (1995) have argued

influential outliers based on examination of Cook's D and DFBETA statistics for each pre- dictor. These two cities have by far the largest Cook's D values and show that the DFBETAs are problematic for a number of variables.

Analyses with unlogged homicide rates yield similar patterns as analyses using logged rates. A few differences are evident in the initial models, but none emerges in tests of our main thesis. In additive models, residential segregation and per- cent black have significant effects on unlogged homicide rates for whites but not on logged rates. In regressions including a quadratic concentrated disadvantage term and an interaction of percent homeowners V concentrated disadvantage, the same patterns hold for both races whether the de- pendent variable is logged or unlogged. Neither term is significant for whites, and both are sig- nificant and in the same direction for blacks.

that social dislocations, including high rates of violence, are rooted in the concentration of disadvantage and community instability. Massey and Denton (1993) contend that ra-


cial residential segregation is an indirect structural cause of high crime because it is a

1 precursor to concentrated disadvantage for blacks (Krivo et al. 1998; Massey and Eggers

1 1990; Massey, Eggers, and Denton 1994; Peterson and Krivo 1999). Blau and Blau (1982) emphasize that high rates of criminal violence are an important cost of social in- equality, especially ascriptive inequality, in contemporary American society. Data for the independent variables come from the 1990 U.S. Census of Population and Housing Summary Tape Files 4A (STF4A)

(U.S. Bureau of the Census 1991), which provide race-specific census tract and city data. We measure concentrated disadvantage with an index combining the concentration of several aspects of disadvantage: poverty, female-headed families, and male jobless- ness. We include three separate P*isolation indices, which represent the probability that:

(1) a poor person has contact in his or her census tract with another poor person; (2) a female-headed family has contact in its cen- sus tract with another female-headed family; and (3) a civilian noninstitutionalized male age 16 and over, who is either unemployed or not in the labor force, has contact in his census tract with another such jobless male. To incorporate intergroup contact between disadvantaged and higher status individuals in our index, we also include a fourth P'hea- sure indicating the probability that a jobless male has contact in his census tract with a person employed in a professional or mana- gerial occupation.

Each P'yndex is calculated separately for blacks and whites as:

When this P*index represents poverty isola- tion, xj is the number of black (or white) poor persons in tract j, xtj is the total number of poor persons in tract j, X is the total num- ber of black (or white) poor persons in the central city, and tj is the total population in tract j. The P* indices for female-headed families, jobless males, and intergroup con- tact are calculated in an analogous fashion. When computing the P* index for female- headed families, tj is the total number of households rather than persons. We con- struct an index (as average z-scores) because of substantial collinearity among the P* measures (Land, McCall, and Cohen 1990).~ Each race-specific z-score is calculated us- ing the mean and standard deviation of the average of the black and white P'\alues to allow for comparisons of absolute levels of the index between the races. If z-scores were calculated relative to the means and standard deviations of the separate black and white P" values, the indices would have means of 0 for both groups even though there are large differences between the races in this variable (Krivo et al. 1998).

To capture community stability, we in- clude homeowner occupancy, operationalized as the race-specific percentage of hous- ing units that are owner-occupied. We mea- sure racial residential segregation with an indicator of residential evenness-the index of dissimilarity (D) for blacks and whites across census track5 The index of dissimi- larity ranges from a low of 0, when blacks and whites are evenly distributed across cen- sus tracts (i.e., every tract has the same per- cent black as the entire city), to a maximum of 100, when blacks and whites are com- pletely segregated (i.e., each tract is either 100 percent black or 100 percent white). Us- ing D allows us to differentiate the indepen- dent effect of the uneven distribution of groups across neighborhoods within a city (segregation) from the effect of the concen- tration of group disadvantage across areas within the city (Peterson and Krivo 1999). The index of dissimilarity value for a city is calculated as:

Race-specific factor analyses confirm that the four indicators represent a single construct. The z-score of the P* for jobless male-professional contact is reverse-coded before it is included in the index.

Following past practice, we exclude tracts dominated by Indian reservations, military bases, hospitals, prisons, or other custodial and residen- tial institutions (Massey and Denton 1987).

where b, and w, are the numbers of blacks and whites, respectively, in tract j, N is the number of tracts in the city, and B and Ware the respective total numbers of blacks and whites in the central city.

An index of interracial socioeconomic in- equality combines (as average z-scores) ra- tios of white to black: median household in- come, percentage of adults who are high school graduates, and percentage of adult


males who are jobless (unemployed or out of the labor force). We control for: (1) per- centage of the black (or white) population that is male and between the ages of 15 and 34; (2) percentage of the total population that is black; (3) region (South or West); and

(4) city population (logged). Population size is log-transformed because crime rates rise at a decreasing rate as population increases (Logan and Messner 1987).


To evaluate the sources of racial differences in the effects of the theoretical predictors of homicide, we first perform race-specific re- gressions to confirm that the influences of the predictors differ between the two groups. We then test whether concentrated disadvan- tage has a curvilinear effect on homicide for each racial group, and whether other theo- retical factors interact with concentrated dis- advantage such that their influences decrease at higher levels of concentrated disadvan- tage.

In analyses of cities, heteroskedasticity is a potential problem because the error vari- ance likely decreases as city population in- creases. We tested for heteroskedasticity by population size within the race-specific models using the Breusch-Pagan test (Greene 1993 :394-95). Analyses indicated significant heteroskedasticity for the regres- sions for blacks but not those for whites. To correct for this, we perform weighted-least- squares (WLS) regressions (with the error variance specified-as inversely proportional to city population size) for blacks; ordinary- least-squares (OLS) techniques are used for the models for whites. The impact of multi- collinearity was assessed using collinearity diagnostics for the OLS models for whites (Belsley, Kuh, and Welsch 1980) and Vari- ance Inflation Factors for the WLS regres- sions for blacks. Although the diagnostics show that the city population parameter for whites is degraded by collinearity, the con- clusions drawn are unaffected-the variable remains significant. For blacks, residential segregation is strongly associated with the other independent variables. However, in models excluding other correlated variables, the standard error for residential segregation is not notably smaller, and hence its signifi- cance is unaffected.

Table 1 presents the means and standard de- viations for all of the variables. These de- scriptive statistics show the common pattern of substantially higher homicide rates for African Americans than for whites. The av- erage logged homicide offending rate per 100,000 is 3.7 for blacks and 1.9 for whites; the mean unlogged rate is more than five times higher for African Americans than for whites, 47.4 versus 9.0. Also, African Americans have much greater concentra- tions of general and specific types of disad- vantage. The largest differences are for pov- erty and female-headed families. Home- ownership rates are dramatically lower for blacks. The two groups have similar percent- ages of young males, although this small dif- ference is significant.

Table 2 presents the regression results pre- dicting homicide rates for African Ameri- cans and whites. The first two columns refer to additive models. The results reveal impor- tant race differences in the magnitude and significance of the predictors of lethal vio- lence. Only concentrated disadvantage, city population, and region are significant for both groups. Two of the four theoretical variables have significantly different effects for whites and blacks: concentrated disad- vantage and percent homeowners. Concen- trated disadvantage has a significant positive influence for both whites and blacks, but the effect for whites is nearly twice as large as that for blacks. The effect of homeownership is 0 for blacks, but negative and large for whites-a 10-percent increase in the percent of homeowners among whites is associated with a 24-percent decrease in the white ho- micide rate (100 x [e-~027X1011). Neither of


the other two theoretical variables, residen-

Table 1. Means and Standard Deviations of Variables: Central Cities, 1990

Variable Whites Blacks

Homicide rate (In)" 1.94 ~72)

Concentrated disadvantage -.84 (.76) P*for poverty .22 (.06) P*for female-headed .15 family (.03)

P* for male joblessness .ll
P*for male joblessness- .14
professional contact (.05)

Percent owner-occupied 55.86 housing (8.70)

Racial residential 58.13 segregation (14.14)

Interracial socioeconomic .02 inequality (.a81

Percent males age 15 to 34 17.63


Percent black 25.49


Region: South


Note: Numbers in parentheses are standard devia- tions; N = 124. a The mean of the unlogged white homicide of- fending rates is 9.0; the mean for blacks is 47.4. Average z-scores for the following four P*indicators of concentrated disadvantage. 7 Difference between the means for whites versus blacks is significant at p .O1 (one-tailed t-tests).

tial segregation and interracial socioeco- nomic inequality, has a significant effect on homicide rates.6 Our results also show racial differences in the influence of some of the control variables-city population and South have significantly stronger effects on lethal violence among whites than blacks.

Intraracial inequality was also examined and was not significant for either racial group.

Table 2. Unstandardized Coefficients from Regressions of Logged Homicide Offending Rates on Independent Variables: Central Cities, 1990

Additive Models Additional Models for Blacks

Independent Variable Concentrated disadvantage index Percent owner-occupied housing Racial residential segregation Interracial socioeconomic inequality Percent males age 15 to 34 Percent black City population (In) South West (Concentrated disadvantage index)2 Concentrated disadvantage index x percent homeowners Constant Adjusted R2

Whites Blacks Model 1 Model 2 Model 3

Note: Numbers in parentheses are standard errors; N = 124. We use weighted-least-squares regressions in the models for Blacks to correct for significant heteroskedasticity by city population size. Buse's R2 values are reported for these equations. Models for whites use ordinary-least-squares regressions.

*p< .05 (two-tailed t-tests)

+p< .05 ++p< .01 (one-tailed t-tests)

'Difference between the coefficients for whites versus blacks significant at p < .01 (one-tailed t-tests)

To explore why the effects of concentrated disadvantage and percent homeowners are much larger for whites than blacks, we ex- amine whether the effects of these charac- teristics weaken as levels of concentrated disadvantage increase. We test two models for each racial group that add, respectively:

(1) a quadratic term for concentrated disad- vantage, and (2) an interaction of percent homeowners and concentrated disadvan- tage.7 For whites, neither term is significant

'We also tested for interactions of concentrated disadvantage with the two other theoreti- cal variables-residential segregation and inter- racial socioeconomic inequality. These interac-

(results not presented). The models for blacks are presented in the last three col- umns of Table 2.The quadratic model shows that concentrated disadvantage among blacks has a significant curvilinear associa- tion with homicide rates; disadvantage has a positive effect at low levels, but its influence weakens at higher level^.^ The interaction of

tions were not significant for whites or blacks.

In additional analyses, we assessed whether there are particular tipping points at which the in- fluence of concentrated disadvantage among blacks changes from notably positive to near zero using a spline function like that used by Crane (1991). These analyses showed a pattern similar to that identified with the quadratic specification

percent homeowners with concentrated dis- advantage is also significant. The effect of percent homeowners is negative and larger at low levels of concentrated disadvantage. The effect levels off in the intermediate range of concentrated disadvantage, and be- comes unexpectedly positive when concen- trated disadvantage is high.

Because both terms are significant, we es- timated a final model that includes both the quadratic and interaction variables. To more clearly illustrate these results, Figure 1 plots predicted homicide rates for blacks at vary- ing levels of concentrated disadvantage (from the 10th [-.69] to the 90th percentiles

[2.08] of observed black values) and percent homeowners (at the mean for blacks [37 per- cent], and one standard deviation below the mean [28 percent] and above the mean [46 percent]). For comparison, predicted homi- cide rates for whites are presented from the 10th (-1.83) to the 90th (.11) percentiles of white concentrated disadvantage (based on the linear model in Table 2). All other inde- pendent variables are held constant at black mean levels for both groups.

Panel A in Figure 1 presents the predicted homicide rates for the full sample of 124 cities. Examining the homicide rates for blacks, in cities with average or above aver- age levels of homeownership, concentrated disadvantage has a sizable positive effect on lethal violence when disadvantage is low. However, the influence of concentrated disadvantage on homicide weakens as dis- advantage levels increase, and is generally small in much of the range observed for Af- rican Americans. For cities in which home- ownership among blacks is low, this curvi- linearity is less pronounced because con- centrated disadvantage has a more modest effect at all levels. In general, the results for blacks are consistent with our predictions- at low levels, concentrated disadvantage has an important effect on homicide, but once levels of concentrated disadvantage are seriously high, community structures are no longer distinguishable enough to produce strong effects on homicide. This curvilinear effect of concentrated disadvan-

in which the effect of black concentrated disad- vantage is large and positive when disadvantage is below .11, and near zero beyond that level.

tage on homicide does not hold for whites because concentrated disadvantage among whites rarely reaches the seriously high lev- els where its effect declines. Only 12 cities (10 percent) have levels of concentrated disadvantage for whites above .11. In sharp

, contrast, 81 percent of cities have levels of concentrated disadvantage for blacks above this point. Are the effects of concentrated disadvan- tage comparable for the two groups when levels for blacks are as low as those ob- served for whites? Indeed, the influence of concentrated disadvantage on black and white homicide is similar within the overlap- ping portions of the two distributions of con- centrated disadvantage (between the vertical lines of Figure 1, Panel A). This is indicated clearly in Table 3, which presents the values of the black slopes for concentrated disad- vantage for this portion of the distribution. These results show that the influence of con- centrated disadvantage among blacks is at least as strong as the linear effect for whites (b = .323 from Table 2) in cities with black homeownership rates at or above their mean level-five of the six coefficients exceed the slope for whites. The most notable compari- son is when the two racial groups have simi- lar levels of both concentrated disadvantage and homeownership. Specifically, home- ownership rates for whites are 46 percent or higher in the vast majority of cities (88 per- cent). For the cities in which rates for blacks reach these levels (and disadvantage is low), the influence of concentrated disadvantage among blacks on homicide is consistently even stronger than that for whites. Clearly, disadvantage is as important for increasing lethal violence among blacks as it is among whites when these racial groups are similarly situated. Panel A in Figure 1 also shows that the in- fluence of homeownership on homicide rates for blacks is more similar to its effect for whites when concentrated disadvantage is low. In the leftmost part of the predicted ho- micide curves for blacks, cities with a 37- percent homeownership rate have lower lev- els of offending than places with a 28-per- cent homeownership rate. Offending de- creases further as percent homeowners for blacks increases to 46 percent. This negative effect of homeownership exists at low levels

Concentrated Disadvantage lndex

Concentrated Disadvantage lndex

Figure 1. Predicted Homicide Offending Rate by Concentrated Disadvantage: Blacks and Whites in Central Cities, 1990

Note: All other independent variables for blacks are held constant at their mean levels. Panel A includes the full sample of central cities. Panel B omits Hartford, Newark, and New York City. Area of overlap between whites and blacks in concentrated disadvantage.

Table 3. Slopes of Concentrated Disadvantage for Blacks at Varying Levels of Low Concentrated Disadvantage and Percent Homeownership: Central Cities, 1990

Percent Homeowners among Blacks

Concentrated Disadvantage 28.0 Percent 36.9 Percent 45.8 Percent

Note: The slope of concentrated disadvantage for whites is .323.

"This represents the 10th percentile for the concentrated disadvantage index for blacks. This represents the 90th percentile for the concentrated disadvantage index for blacks.

of concentrated disadvantage but becomes smaller as concentrated disadvantage in- creases. To illustrate, the slope for percent homeowners is -.014 at the 10th percentile of concentrated disadvantage for blacks (from the final model of Table 2).This effect is just over half that for whites (b, = -.027, Table 2). Still, it is negative and more simi- lar to the slope for whites than in the addi- tive model (bb= -.000, Table 2). Homeownership's influence reduces to near zero for a notable portion of the concentrated dis- advantage distribution, but turns positive at very high levels of concentrated disadvan- tage. This result is unexpected. Further ex- ploration reveals that it is due to three cities that have exceptionally low homeownership combined with extremely high concentrated disadvantage for blacks: h art ford, Connecticut, Newark, New Jersey, and New York City. These places also have somewhat less lethal violence among blacks than other highly disadvantaged cities. When these three cities are removed from the analysis (Panel B of Figure I), the positive effect of homeownership is essentially eliminated and generally equals zero in all places with very high concentrated disadvantage among black^.^

Major theories of crime and violence lead one to expect similar etiological processes for different racial groups. Yet a common

Excluding Hartford, Newark, and New York City does not alter the significance or pattern of any of the other relationships in the models pre- sented in Table 2.

finding is that some crime-generating con- ditions have stronger effects for whites than for blacks. We have sought to account for these racial differences. We argued that varying effects of predictors may result from the dramatically different social posi- tions of blacks and whites in U.S. cities. In particular, the high level and high concen- tration of disadvantage in the African American population may create a situation that actually reduces the importance of these very conditions for generating higher homi- cide rates for blacks. Variation in structural disadvantage at very high levels is unlikely to produce communities that are qualita- tively different from one another. Hence, criminal violence should not be systemati- cally associated with variation in structural conditions for African Americans. By con- trast, whites generally live in communities with a lower prevalence of violence-produc- ing factors. Thus for whites, variation in lev- els of key theoretical variables should re- flect more meaningful differentiation across communities, and hence have stronger ef- fects. Taken together, these arguments im- ply that: (1) Important structural causes of homicide will have relatively weak effects in highly disadvantaged social contexts; and

(2) the differing black and white effects fohnd in previous studies reflect the fact that these racial groups fall predominantly in separate portions bf the distribution of-dis- advantage. Accordingly, in places where the circumstances of blacks and whites are simi- lar, crime-generating social conditions should have comparable effects on homi- cide.

Our analyses of the race-specific determi- nants of homicide offending rates for U.S.

cities in 1990 provide some support for these ideas. First, we find that racial differences in the homicide-generating processes are siz- able for two key theoretical determinants: concentrated disadvantage and residential stability (percent homeowners). These fac- tors have significantly larger effects on white homicide rates than on black rates. Second, the differences in effects appear to be rooted in the varying social positions of the two races. This is seen in the leveling off of the influence of concentrated disadvan- tage for blacks at the high levels that pre- dominate among African Americans.

In addition, there is a strikingly strong positive impact of black concentrated disad- vantage on homicide when levels of concen- trated disadvantage are comparable for whites and blacks. Support for our proposed explanation is further demonstrated in the significant interaction between homeowner- ship rates and concentrated disadvantage for blacks, whereby homeownership has a nega- tive effect on homicide only when concen- trated disadvantage is relatively low. In con- trast, neither the curvilinear nor interactive effects of concentrated disadvantage are sig- nificant for whites. This is because the white population in most cities has a low enough level of concentrated disadvantage that this factor has pronounced effects. Overall, these findings imply that if blacks and whites held similar positions in relation to structural dis- advantage, differences in criminogenic fac- tors would operate similarly for the two groups.

Still, our evidence is not completely de- finitive because of the strong connection be- tween race and disadvantage in the United States. As our data indicate, blacks and whites seldom have comparable levels of disadvantage. Indeed, racial differences in disadvantage are so great that it is impos- sible to assess what the effects for whites would be if they were as disadvantaged as the average African American in most urban areas. Obviously, the leveling-off effect of disadvantage that we observe for blacks might, or might not, be evidenced for whites; which possibility is correct is likely to re- main a matter of speculation given the dra- matic racial inequality embedded in U.S. so- ciety. As such, the empirical implications for current research are clear. For the time be- ing, it is imperative that models of crime, in- cluding homicide, be explored separately for blacks and whites because the similarity of conditions required for combining groups (observing uniform effects) do not exist in the vast majority of places.

Despite some illuminating findings, this research reveals a large unexplained racial difference in homicide rates (see Figure 1). Clearly, new thinking and empirical analy- ses are required to gain a fuller understand- ing of the sources of this sizable race differ- ential. Future investigations should consider additional aspects of disadvantage that could affect homicide. We have been guided by re- cent social disorganization literature that fo- cuses on the consequences of poverty, job- lessness, female-headed households, and the absence of professional workers. Studies should also explore factors such as labor market conditions (low wages and second- ary-sector jobs) and high school dropout rates to determine whether their effects are similar to those found here for the index of concentrated disadvantage.

In addition to considering different aspects of deprivation, studies should consider how the broader meaning of race af- fects the specific position of blacks and whites vis-a-vis crime and violence in con- temporary American society. For example, the social disorganization perspective emphasizes the role of informal networks in facilitating social control in communities. Here, we included factors that are purported to produce these networks (economic depri- vation and instability), but we did not mea- sure networks directly or explore the possi- bility that such networks may have different manifestations in black and white commu- nities. For example, Patti110 (1998) points out that in contrast to white neighborhoods, black middle-class areas have higher levels of poverty and are in closer geographic proximity to poor and high-crime neighbor- hoods. As such, the informal networks that provide the underlying social organization of these communities include long-term residents who are part of groups that actu- ally contribute to or draw crime and vio- lence to their neighborhoods (e.g., gang members and leaders, persons involved in drug markets, etc.). Also, because of the greater spatial clustering of disadvantage among blacks than among whites, the insti- tutional and economic resources that dimin- ish the likelihood of violent crime are fewer and farther away for black communities than is the case for disadvantaged white neighborhoods. These differential manifes- tations of social organization and disadvan- tage in black and white communities may in part account for the racial gap in homicide, and thus, should be examined in future re- search.

In the meantime, our findings support the widely held view that some of the major structural causes of violent crime are invari- ant across groups. The crime-generating pro- cesses considered here do not differ much between African Americans and whites when the two racial groups are similarly situated.

However, such a finding must be interpreted within the context of contemporary U.S. society in which racial similarity of conditions is rare. Blacks and whites do not live in com- parable community settings in most Ameri- can cities. In only 19 percent of the places in our sample (24 cities) do blacks experi- ence as low a level of concentrated disad- vantage as is experienced by the vast major- ity of whites (112 cities), and these are the very places in which few blacks reside. Only

4.2 percent of African Americans in the 124 cities examined here live in this subset of cities. This means that only a small number of African Americans actually experience the same conditions and processes that oper- ate for whites. In the remaining cities where black disadvantage is more highly concen- trated, and where over 95 percent of blacks in our sample reside, homicide is more weakly impacted by important causal mechanisms and hence appears more intran- sigent.

Lauren J. Krivo is an Associate Professor of So- ciology at Ohio State University. Her current re- search interests include black-white differences in various types of homicide, neighborhood dis- advantage and youth and adult crime, and sex differences in the English-language proficiency of immigrants. With Ruth Peterson, she is begin- ning a national study of urban neighborhood crime patterns.

Ruth D. Peterson is Professor of Sociology and the Director of the Criminal Justice Research Center at Ohio State University. Her major re- search interests include social disadvantage and black-white differences in homicide, neighbor- hood crime patterns, legal decision-making and sentencing, and crime and deterrence.

Belsley, David A,, Edwin Kuh, and Roy E. Welsch. 1980. Regression Diagnostics: Iden- tifying Influential Data and Sources of Col- linearity. New York: John Wiley and Sons.

Berry, Brian J. L. 1991. Long-Wave Rhythms in Economic Development and Political Behav- ior. Baltimore, MD: Johns Hopkins University Press.

Blau, Judith R. and Peter M. Blau. 1982. "The Cost of Inequality: Metropolitan Structure and Violent Crime." American Sociological Re- view 47: 1 14-29.

Center for the Study and Prevention of Violence. 1995.Estimated Race-Specific Homicide Rates Calculated from the Supplementaly Homicide Files for 1979-1981 and 1989-1991. Boulder, CO: Center for the Study and Prevention of Violence.

Crane, Jonathan. 1991. "The Epidemic Theory of Ghettos and Neighborhood Effects on Drop- ping Out and Teenage Childbearing." American Journal of Sociology 96: 1226-59.

Granovetter, Mark. 1978. "Threshold Models of Collective Behavior." American Journal of So- ciology 83: 1420-43.

Greene, William H. 1993. Econometric Analysis. 2d ed. Englewood Cliffs, NJ: Prentice Hall.

Harer, Miles D. and Darrell Steffensmeier. 1992. "The Differing Effects of Economic Inequal- ity on Black and White Rates of Violence." Social Forces 70:1035-54.

Krivo, Lauren J. and Ruth D. Peterson. 1996. "Extremely Disadvantaged Neighborhoods and Urban Crime." Social Forces 75:619-50.

Krivo, Lauren J., Ruth D. Peterson, Helen Rizzo, and John R. Reynolds. 1998. "Race, Segrega- tion, and the Concentration of Disadvantage: 1980-1990." Social Problems 45:61-80.

LaFree, Gary. 1999. "Declining Violent Crime Rates in the 1990s: Predicting Crime Booms and Busts." Annual Review of Sociology 25: 145-68.

LaFree, Gary and Kriss A. Drass. 1996. "The Ef- fect of Changes in Intraracial Income Inequal- ity and Educational Attainment on Changes in Arrest Rates for African Americans and Whites, 1957 to 1990." American Sociological Review 61:614-34.

LaFree, Gary, Kriss A. Drass, and Patrick O'Day. 1992. "Race and Crime in Postwar America: Determinants of African-American and White Rates, 1957-1988." Criminology 30:157-88.

Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1990. "Structural Co- variates of Homicide Rates: Are There Any In- variances across Time and Social Space?" American Journal of Sociology 95:922-63.

Logan, John R. and Steven F. Messner. 1987. "Racial Residential Segregation and Suburban Violent Crime." Social Science Quarterly 68:510-27.

Massey, Douglas S. and Nancy A. Denton. 1987. "Trends in the Residential Segregation of Blacks, Hispanics, and Asians: 1970-1980." American Sociological Review 52:802-25.

. 1993. American Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.

Massey, Douglas S. and Mitchell L. Eggers. 1990. "The Ecology of Inequality: Minorities and the Concentration of Poverty, 19701980." American Journal of Sociology 95: 1153-88.

Massey, Douglas S., Mitchell L. Eggers, and Nancy A. Denton. 1994. "Disentangling the Causes of Concentrated Urban Poverty." International Journal of Group Tensions 24:267


Messner, Steven F. and Reid M. Golden. 1992. "Racial Inequality and Racially Disaggregated Homicide Rates: An Assessment of Alterna- tive Theoretical Explanations." Criminology 30:421-47.

Parker, Karen F. and Patricia L. McCall. 1999. "Structural Conditions and Racial Homicide Patterns: A Look at the Multiple Disadvan- tages in Urban Areas." Criminology 37:447


Pattillo, Mary E. 1998. "Sweet Mothers and Gangbangers: Managing Crime in a Black Middle-Class Neighborhood." Social Forces 76:747-74.

Peterson, Ruth D. and Lauren J. Krivo. 1993. "Racial Segregation and Black Urban Homi- cide." Social Forces 71:1001-26.

. 1999. "Racial Segregation, the Concen- tration of Disadvantage, and Black and White Homicide Victimization." Sociological Forum 14:495-523.

Phillips, Julie A. 1997. "Variation in African- American Homicide Rates: An Assessment of Potential Explanations." Criminology 35:527


Quercia, Roberto G. and George C. Galster. 1999. "Threshold Effects and Neighborhood Change." Report to the Annie E. Casey Foun- dation. Baltimore, MD: Annie E. Casey Foun- dation.

Riedel, Marc. 1990. "Nationwide Homicide Data Sets: An Evaluation of the Uniform Crime Re- ports and the National Center for Health Sta- tistics Data." Pp. 175-205 in Measuring Crime: Large-Scale, Long-Range Efforts, edited by D. L. MacKenzie, P. J. Baunach, and

R. R. Roberg. Albany, NY: State University of New York Press.

Sampson, Robert J. 1987. "Urban Black Vio- lence: The Effect of Male Joblessness and Family Disruption." American Journal of So- ciology 93:348-82.

Sampson, Robert J. and William Julius Wilson. 1995. "Race, Crime, and Urban Inequality." Pp. 37-54 in Crime and Inequality, edited by

J. Hagan and R. D. Peterson. Stanford, CA: Stanford University Press.

Schelling, Thomas C. 1971. "Dynamic Models of Segregation." The Journal of Mathematical Sociology 1:143-86.

Shihadeh, Edward S. and Nicole Flynn. 1996. "Segregation and Crime: The Effect of Black Social Isolation on the Rates of Black Urban Violence." Social Forces 74:1325-52.

Shihadeh, Edward S. and Graham C. Ousey. 1996. "Metropolitan Expansion and Black So- cial Dislocation: The Link between Suburban- ization and Center-City Crime." Social Forces 75:649-66.

, 1998. "Industrial Restructuring and Vio- lence: The Link between Entry-Level Jobs, Economic Deprivation, and Black and White Homicide." Social Forces 77: 185-206.

Shihadeh, Edward S. and Darrell J. Steffens- meier. 1994. "Economic Inequality, Family Disruption, and Urban Black Violence: Cities as Units of Stratification and Social Control." Social Forces 73:729-5 1.

    Bureau of the Census. 1991. 1990 Census of Population and Housing: Summary Tape File 4A (Ohio) [MRDF]. Washington, DC: U.S. Bureau of the Census [producertdistribu- tor].

    Department of Justice. 1999. Sourcebook of Criminal Justice Statistics-1998. Bureau of Justice Statistics. Washington, DC: U.S. Gov- ernment Printing Office.

Williams, Kirk R. and Robert R. Flewelling. 1987. "Family, Acquaintance, and Stranger Homicide: Alternative Procedures for Rate Calculations." Criminology 25:543-60. , 1988. "The Social Production of Crimi- nal Homicide: A Comparative Study of Disag- gregated Rates in American Cities." American Sociological Review 53:421-3 1.

Wilson, William Julius. 1987. The Truly Disad- vantaged: The Inner City, the Underclass, and Public Policy. Chicago, IL: University of Chi- cago Press.

. 1996. When Work Disappears: The World of the New Urban Poor. New York: Alfred A. Knopf.

  • Recommend Us