#1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

Moving Beyond Poverty: Neighborhood Structure,Social Processes, and Health*

CHRISTOPHER R. BROWNING The Ohio State University KATHLEEN A. CAGNEY University of Chicago

Delivered by Ingenta to Journal of Health and Social BehaviorWelch 2003, V Medicalol 44 (December): Library 552– (cid571 82007175), We investigate the impact of neighborhoodJohns Hopkins structural University characteristics, (cid 90000553) social organization, and on self-ratedIP: 127.0.0.1 health in a large, cross-sectional sample of urban adults. Findings indicateDate: that 2005..04..18..17..00..neighborhood affluence is a more pow- erful predictor of health status than poverty, above and beyond individual demographic background, socioeconomic status, health behaviors, and insur- ance coverage. Moreover, neighborhood affluence and residential stability interact in their association with health. When the prevalence of affluence is low, residential stability is negatively associated with health. Neighborhood affluence also accounts for a substantial proportion of the racial gap in health status. Finally, is a significant positive predictor of health but does not mediate the effects of structural factors.

Research on the contribution of neighbor- these efforts, neighborhood effects research on hood characteristics to individual health has health has typically focused on only one progressed rapidly over the last decade. dimension of neighborh ood structure— the Mounting empirical evidence of neighborhood prevalence of economic deprivation—and has socioeconomic structure (SES) effects on a yet to explore competing hypotheses regarding range of adult outcomes including mortality the community level mediators of structural (Haan, Kaplan, and Camacho 1987), heart dis- effects on health. ease (LeClere, Rogers, and Peters 1998), num- We draw on social disorganization and col- ber of chronic conditions (Robert 1998), and lective efficacy theory (Shaw & McKay 1969; self-rated health (Malmstrom, Sundquist, and Sampson, Raudenbush, and Earls 1997), Johansson 1999) has fueled calls for continued Wilson’s theory of neighbo rhood decline research on the health effects of multiple (Wilson 1987; 1996), and research exploring dimensions of community SES and the mech- the health effects of neighborhood context to anisms that may account for the community elaborate and test two sets of hypotheses. First, structure-health link (Robert 1999). Despite we extend the typical focus on the health con- sequences of neighborhood poverty to include * Paper presented in part at the 2001 meetings of the a range of other structural characteristics of Population Association of America, Washington, D.C. Support for this research was provided by the neighborhoods including the concentration of Ohio State Seed Grand Initiative. Thanks to Felicia affluence, residential stability, and ethnic het- LeClere, Lauren Krivo, Ming Wen, and participants erogeneity. Our focus on the role of affluence, in the University of Chicago Population Research in particular, builds on recent theory and Center W orkshop for their helpful research highlighting the potentially important comments. W e wish to thank Robert Sampson, role of middle and upper-middle class resi- Felton Earls, and members of the Project on Human dents in anchoring community and Development in Chicago neighborhoods for gener- ously providing access to the Community providing organizational and economi c data. W e also thank the Metropolitan Community resources with which to promote a healthy Information Center for providing access to the environment (Cagney, Browning, and W en Metro Survey data. 2002; Robert 1998). Wilson’s theory of social

552 #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 553 isolation adds the additional insight that resi- basis for this association has received compar- dential stability may interact with socioeco- atively less attention. We draw on recent exten- nomic context in influencing the well-being of sions of the longstanding tradition of theory urban residents (Wilson 1987; 1996; Ross, and research on the role of “social disorgani- Reynolds, and Geis 2000). zation” in the spread of urban social problems Second, we elaborate and test a number of (Bursik 1988; Shaw & McKay 1969; hypotheses regarding the community -level Sampson, Raudenbush, and Earls 1997). mechanisms thought to mediate the effects of Though originally developed as an explanation structural conditio ns at the neighbo rhood for varying rates across urban neighbor- level. The mechanisms we consider include hoods, this body of theory identifies specif ic neighborhood level social support and socia- dimensions of neighborhood structure that bility (Berkman 1995; Berkman, Delivered Glass, by mayIngenta be relevant to to the distribution of a wide Brissette, and Seeman 2000; Berkman and range of phenomena, including poor health. Syme 1979; Hawe and Shiell 2000)Welch, normative MedicalSocial Library disorganization (cid 82007175), theory has a lengthy transmission of health-compromisingJohns behavior Hopkinshistory University in urban (cid . 90000553) In Shaw and (Fitzpatrick and Lagory 2000) , IP:and 127.0.0.1 health- McKay’s (1969) initial statements of the related collective eff icacy (BrowningDate: 2005..04..18..17..00.. and approach and subseque nt elaborat ions Cagney 2002; Sampson, Raudenbush , and (Kornhauser 1978), features of neighborhood Earls 1997). Though some studies have exam- structure—principa lly poverty, resident ial ined the impact of individual mechanisms, this instability, and ethnic heterogenei ty—were research has either focused on relatively large hypothesized to limit resources with which to units of aggregatio n (e.g., the state level support local institutions, diminish community [Kawachi, Kennedy , Lochner, and Prothrow- attachments through high population turnover, Stith 1997]) or has used person level opera- and impede communicatio n across diverse tionalizations of processes specified to operate racial/ethnic groups. In turn, these conditions at the neighborhood level (Ross and Mirowsky result in attenuated community-level social 2001). In contrast , we consider multiple control of crime. In their empirical analyses, hypotheses regarding the neighborhood level Shaw and McKay found that these structural mediators of socioeconomic structure effects factors continued to affect crime rates regard- on health simultaneously. less of ethnic and racial population succession, We investigate neighborho od effects on suggesting that macro level processes operate health using a multilevel statistical approach independently of the characteristics of individ- (Raudenbush and Bryk 2002; Snijders and uals who comprise disadvantaged neighbor- Bosker 1999). Three data sources are hoods. This insight remains critical to research employed: 1990 decennial census data, the on neighborhood effects today . For instance, 1994–95 Project on Human Development in deficits in the health of African-Americans and Chicago Neighborhoods Community Survey Latinos by comparison with whites observed at (PHDCN-CS), and the 1991– 2000 Metro- the individual level may actually be capturing politan Community Information Center-Metro inequality in the social environments these Survey (MCIC-MS)—a serial cross-section of groups typically occupy. Chicago adults. The combination of these data These early efforts at describing the spatial sets offers a unique opportunity to assess the organization of social problems have seen contribution of multiple dimensions of neigh- renewed interest over the last two decades. borhood structure and social process to varia- Research examining the contextual impact of tions in the health of urban residents. economic conditions on health, for instance, has highlighted the potential relevance of social disorganization theory to health out- THEORETICAL PERSPECTIVES comes (Robert 1999). Indeed, the neighbor- hood level consequences hypothesized to fol- Neighborhood Socioeconomic Structure and low from structural deficits in the disorganiza- Health tion approach —attenuated institut ional strength, limited network interaction, dimin- The contribution of neighborhood structure ished neighborhood attachment, and low levels to health has been the subject of extensive of informal (Kasarda and empirical investigation, but the theoretical Janowitz 1974; Kornhauser 1978; Sampson #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 554 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR 1997)—can be seen as potentially important proportion of households with incomes over determinants of health as well as outcomes $50,000 was significantly associated with bet- such as victimization. We describe this process ter health independent of demographic and in more detail in our discussion of the hypoth- health background while the prevalence of esized mechanis ms linking neighborhood poverty was not a significant predictor of structure with health. In particular, we empha- health. Robert (1999) found that a measure of size the role of collective efficacy— or the the proporti on of residents with incomes combination of social cohesion and shared greater than $30,000 had a direct positive expectations for beneficial action on behalf of effect on the health of individu als. the communi ty—as a critical interven ing Consequently, a focus on the prevalence of social process relevant to health outcomes affluence as an important structural determi- (Sampson, Raudenbush, and EarlsDelivered 1997). by nantIngenta of health to is theoretically plausible and An important insight of the collective effi- consistent with previous empirical findings. cacy extension of social disorganizationWelch theory Medical Library (cid 82007175), Johns Hopkins FinallyUniversity, Wilson’ (cids theory 90000553) of social isolation concerns the potential significance of the pres- (1987; 1996) has called attention to the poten- ence of affluent residents for neighborhood IP: 127.0.0.1 Date: 2005..04..18..17..00..tial for structural features of communities to social organization (Sampson, Morenoff, and interact in their influence on individual well- Earls 1999; Sampson, Raudenbush, and Earls being. Wilson’s model emphasizes a two-stage 1997). Consistent with this emphasis, research dynamic process of neighborhood decline. on health has increasingly emphasized the dis- First, the flight of middle class residents of tribution of advantage as a critical dimension inner city communities results in the declining of community economic structure (Cagney , viability of local institut ions—includin g Browning, and W en 2002; Robert 1998). churches, schools, voluntary organization s, Affluent residents may draw more and higher and the family—and corresponding deficits in quality health services and reinforce the viabil- the capacity of residents to maintain informal ity of local organizations and services that social controls. Second, as the regulatory indirectly benefit health (e.g, parks and recre- capacity of the community diminishes, resi- ational services ). Indeed, while f indings dents are increasingly exposed to problem regarding the negative effects of disadvantaged behaviors. In turn, these (potentially health- neighborhood SES on health have been rea- compromising) behavioral orientations may be sonably consistent (Davey Smith, Hart, Watt, Hole, and Hawthome 1998; Diez-Roux, Nieto, transmitted by precept, further contributing to Muntaner, Tyroler, Comstock, Shahar, Cooper, community decline. The structural context in Watson, and Szklo 1997; Haan, Kaplan, and which this process occurs most eff iciently is Camacho 1987; Humphreys and Carr-Hill one characterized by both widespread econom- 1991; Jones and Duncan 1995; Katz, Kling, ic disadvantage and constrained mobility. and Liebman 2001; Krieger 1992; LeClere, Contrary to the traditional social disorgani- Rogers, and Peters 1997; Leventhal and zation emphasis on the beneficial effects of Brooks-Gunn 2001; Waitzman and Smith residential stability, then, this approach views 1998; Y en and Kaplan 1999), some studies residentially stability as potentially exacerbat- have found no effect of concentrated poverty ing the effects of economic disadvantage. We or disadvantage after controlling for individual extend investigation of the consequences of level SES (Ecob 1996; Sloggett and Joshi social isolation to consider the potential inter- 1994). Moreover, studies that find an associa- action between stability and both the preva- tion between neighborhood SES and health lence of poverty and affluence. Consistent with have often employed measures that tap some the second stage of Wilson’s model of neigh- combination of the prevalence of poverty and borhood decline, stable poverty may encourage affluence (LeClere, Rogers, and Peters 1998; the dissemina tion of health-co mpromising Ross and Mirowsky 2001). subcultures (Fitzpatrick and LaGory 2000). Few studies have specifically examined the Consistent with the f irst stage of Wilson’ s effects of the upper end of the neighborhood model, however, stable neighborhoods that socioeconomic continuum on health. In an lack affluent residents may experience long investigation of community SES effects on the term declines in the institut ional fabric health of older persons, Cagney , Browning, (including, potentially, the quality and avail- and Wen (2002) found that a measure of the ability of health services) and informal social #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 555 control capacity of the community, with conse- gest that networks of networks at the commu- quences for health. nity level influence egocentric connectedness In sum, building on recent extensions to the and health-enhancing processes such as social social disorganization approach in the form of support (House 1981), positive influence, and collective efficacy theory and Wilson’s con- sociability (Berkman, Glass, Brissette, and cept of social isolation, we focus on aspects of Seeman 2000; Rook 1990). Evidence of the neighborhood socioeconomic structure beyond positive effects of social support on health is poverty that are theoretically relevant to indi- powerful and consistent (Berkman and Syme vidual health. These include concentra ted 1979; House, Robbins, and Metzner 1982; affluence, residential stability, and ethnic het- Orth-Gomer and Johnson 1987), suggesting erogeneity. Acknowledging the unique condi- that the prevalence of social support activity tions of extremely disadvantagedDelivered and socially by andIngenta community to social engagement at the isolated communities, we also consider the neighborhood level may also be relevant for potential for community economic Welch structure Medicalhealth Library outcomes. (cid 82007175), (both the concentration of poverty Johns and afflu- Hopkins SocialUniversity Organization, (cid 90000553) Subcultural Transmis- ence) to condition the impact of residentialIP: 127.0.0.1 sta- sion and Health. The subcultural perspective bility on health outcomes. A focusDate: on structur- 2005..04..18..17..00..on health can be characterized with reference al characteristics of communities alone, how- to the second stage of Wilson’ s theory of ever, leaves unanswered a critical question: neighborhood decline. According to Wilson, what mechanisms link community structural the combination of poverty and immobility characteristics with health outcomes? results in the social and spatial isolation of neighborhood resident s from mainstream sources of influence. Modeling and transmis- Mechanisms Linking Neighborhood Structure sion of problema tic behaviora l strategie s with Health ensue. Some of these behavioral orientations may be adaptive for survival (such as display Hypotheses regarding the impact of distinct of a “ tough” or violent demeanor [Anderson dimensions of community structure require 1990] ). However, cultural transmission of vio- specification of the mechanisms by which lent and other risky behavior (such as smoking, individual health status is likely to be influ- drinking, risky sexual activity, and poor diet) enced. A focus on theories of social organiza- in these contexts may have serious conse- tion points largely to the intervening impact of quences for health. the prevalence of social ties and collective effi- Applying Wilson’ s concept of social isola- cacy, as we shall see. However, other plausible tion explicitly to health outcomes, Fitzpatrick sociocultural mechanisms have been hypothe- and LaGory (2000) suggest that disadvantaged sized to channel the impact of community communities with limited access to extra-local SES—most notably, subcultural tolerance of mainstream institutions may experience the health compromising behavior and anomie at emergence of “ health-related subculture s.” the neighborhood level. Below , we consider Two aspects of these subcultural orientations hypotheses regarding the effects of social may have consequences for health: 1) toler- interaction and sociability, health-related sub- ance for risky lifestyles and 2) anomie or , and collective efficacy in turn. detachment from conventional values. First, Informal Social Support, Sociability, and widespread community level tolerance of risk Health. The increasingly vast literature on the behavior, including smoking, drinking, drug role of informal social supports in promoting use, and poor diet, may facilitate the spread of health at the individual or egocentric network such behavior within disadvantaged contexts, level has fostered interest in the effects of com- with consequences for health. munity level social network characteristics on Second, structural disadvantage, character- health. Structurally disadvantaged neighbor- ized by widespread poverty, lack of access to hoods may be less capable of sustaining viable employment, and bleak economic prospects, social networks and may suffer from deficits in may lead to anomic social conditio ns local social support and sociability (Berkman (Durkheim 1979; Sampson and Bartusch and Breslow 1983). 1998) in which neighborhood residents ques- Berkman and colleagues (Berkman, Glass, tion basic normative orientations (e.g., the Brissette, and Seeman 2000), for instance, sug- value of abiding by the law and of employing #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 556 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR conventional means such as education and ed the powerful effects of collective efficacy hard work to achieve success). Since the bene- on rates of violence— suggesting that health fits of adherence to basic cultural values are may be influenced by high levels of collective perceived not to accrue to residents of some efficacy through limiting the health damaging disadvantaged neighborhood s, the force of consequences of violent victimization. Other value based prescriptions weakens in these forms of problem behavior including illicit contexts. Possible adaptations to these condi- substance use, alcohol abuse, child and elder tions include the emergence of largely short- neglect/abuse, and reckless behavior may also term, instrumental orientations toward goals be held in check by high levels of collective marked by little concern for the future. As social control (Sampson 1997). Second, collec- applied to health, this hypothesis would sug- tive efficacy may enhance the capacity of com- gest that anomie-induced short-term, Delivered “ here by munitiesIngenta to to attract and maintain high quality and now” orientations emphasizing satisfac- health services and amenities such as commu- tion of immediate needs will tend Welchto be associ- Medicalnity Library health clinics (cid 82007175),and safe recreational space. ated with health-c ompromisingJohns behaviors. HopkinsThird, University collective (cid efficacy 90000553) may aid in correcting While an extensive literature has IP:documented 127.0.0.1 or avoiding the accumulation of neighborhood the role of individual level risk Date:behaviors 2005..04..18..17..00.. in physical hazards such as decaying infrastruc- promoting individual level health (Blair, Kohl, ture and housing stock. Communities with the Paffenbarger, Clark, Cooper, and Gibbons capacity to solicit and secure external 1989; Paffenbarger, Hyde, Wing, Lee, Jung, resources to correct potentially risky condi- and Kampert 1993), few studies to date have tions and monitor vulnerable residents (e.g., attempted to assess the health effects of com- the elderly) are likely to enhance health. munity-level tolerance of risk behavior and Finally, the effect of widespread trust and anomie. neighborhood attachment on factors such as Collective Efficacy and Health. A third per- fear and self-respect may improve the health spective highlights the capacity for action on and well-being of residents— even if they do behalf of community goals as the critical inter- not benef it from direct network support vening mechanism linking community struc- (Kawachi and Berkman 2000). ture with health. Sampson and colleagues The empirical investigation to follow con- (Sampson, Raudenbush, and Earls 1997) have siders the impact of affluence, residential sta- encapsulated this process in the concept of col- bility, and ethnic heterogeneity on health, in lective eff icacy, which emphasizes mutual addition to poverty, in an effort to extend the trust and (social cohesion), and typically more narrow focus on the latter in shared expectat ions for pro-socia l action analyses of neighborhood effects on health. We (informal social control) in theorizing the then consider the potential for mediating or impact of neighborhood social organization on unique effects of the three intervening social local resident’s well-being. While acknowledg- processes considered—social support and ing that local social ties may contribute to sociability, subcult ural toleranc e for risk these dimensions of community social organi- behavior and anomie, and collective efficacy . zation, the collective eff icacy approach must We stress that these social processes need not be seen as distinct from the neighborhood be considered mutually exclusive or funda- social support and sociability perspective to mentally contradi ctory. Neverthel ess, by the extent that it emphasizes the sense of attempting to adjudicate between these mecha- attachment to community and the perceived nisms, we take seriously their distinct willingness of community residents to inter- emphases and, in so doing, hopefully illumi- vene on each others’behalf regardless of pre- nate the specific community level social con- existing social ties .1 ditions that contribute to individual health. The pathways through which neighborhood collective eff icacy may influenc e health include the social control of health-risk behav- DATA AND MEASURES ior, access to services and amenities, the man- agement of neighborhood physical hazards, We use three data sources to explore the and psychosocial processes (Browning and association between neighborhood context and Cagney 2002; Kawachi and Berkman 2000). health—the 1990 Decennia l Census, the First, Sampson, et al. (1997) have demonstrat- 1994–95 Project on Human Development in #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 557 Chicago Neighborhoods-Commu nity Survey Strawbridge, Camacho, and Cohen 1993) and (PHDCN-CS), and the 1991–2000 Metropoli- health care utilizat ion (Andersen 1987; tan Chicago Information Center Metro Survey Malmstrom, Sundquist, and Johansson 1999) (MCIC-MS). Measures of neighborhood struc- has been widely documented. Furthermore, tural disadvantage are taken from census data. self-assessed health is a stronger predictor of Measures of neighborhood social processes are mortality than is physician-assessed health constructed from the PHDCN-CS, a probabili- (Mossey & Shapiro, 1982). Although validity ty sample of 8,782 residents of Chicago focus- assessments of the self-rated health measure ing on respondent assessments of the commu- across dimensions such as gender, race and nities in which they live. The PHDCN-CS ethnicity still merit further exploration (Idler combined 865 census tracts into 343 “neigh- and Benyamini 1997), initial investigations borhood clusters” (NCs) maintainingDelivered relative by indicateIngenta that to its predictive capacity is compara- population homogene ity with respect to ble for Latinos, African Americans and Whites racial/ethnic, socioecon omic, housing, Welch Medicaland (Finch, Library Hummer, (cid 82007175),Reindel, and V ega 2002; family structure characteristics (NCs Johns average HopkinsGibson University 1991; Johnson (cid 90000553) and Wolinsky 1994). 5 roughly 8,000 people). 2 The three-stage IP: 127.0.0.1 sam- The MCIC-MS asks “In general, would you pling strategy selected city blocksDate: within 2005..04..18..17..00..NCs, say your health is: excellent, good, fair or dwelling units within blocks, and respondents poor?” Consistent with a number of previous (one adult, age 18 or over, per household) analyses (Idler and Benyamini 1997; within dwelling units. The PHDCN-CS sam- Benyamini and Idler 1999), we treat self-rated pling strategy was intended to capture a with- health as a dichotomy, collapsing reports of in-NC sample sufficient to estimate neighbor- “fair” or “ poor” health in order to capture hood characteristics based on aggregated indi- respondents at the low end of the health distri- vidual level data— ensuring the reliability of bution (W olinsky and Tierney 1998). Table 1 neighborhood level measures of social reports the proportion of MCIC-MS respon- processes.3 dents reporting each level of health; 22.3% The dependent variable and individual level reported being in fair or poor health. predictors are drawn from the MCIC-MS. The MCIC-MS is a serial cross-section of adults ages 18 and older residing in the six county Independent Measures metropolitan Chicago area (averaging around 3,000 respondents per wave, roughly 1,000 of We constructed measures of neighborhood which fall within the City of Chicago) focus- level social composition from the 1990 decen- ing on a wide variety of individual level mea- nial census. In operation alizing economic sures of well-bein g, includi ng self-rated structure, we chose to focus on the prevalence health.4 The MCIC-MS sample captured 339 of poor and middle/ upper middle class resi- of the 343 PHDCN NCs (N = 3,272). Though dents as defined by income. For the purposes the data are not longitudinal, rendering the of a comparative analysis, this approach allows determination of causation problematic, they for an assessment of concentrations in the do include a number of individual level health upper and lower tails of a single dimension. 6 background controls (detailed below). Poverty was operationalized as the proportion of neighborhood residents with incomes below the poverty line. Affluence captured the per- Dependent Measure centage of households with incomes over $50,000. A r esidential stability scale was con- Our dependent variable is a measure of self- structed based on scores from a factor analysis rated health, considered a robust measure of of measures of housing tenure (percent living general health status (Goldstein, Siegel, and in the same house since at least 1985) and the Boyer 1984; Wilson and Kaplan 1995). The percent of housing occupied by owners (factor validity of self-rated health as a predictor of loadings exceeded .75). 7 Tapping ethnic het- mortality (Benyamini and Idler 1999; Idler and erogeneity in the context of Chicago, a second Angel 1990; Idler and Benyamini 1997; factor—immigrant concentration —was marked Kaplan, Barell, and Lusky 1988), morbidity by high factor loadings for percent Latino (.88) (Ferraro, Farmer, and Wybraniec 1997), subse- and percent foreign born (.70). 8 quent disability (Idler and Kasl 1995; Kaplan, Collective eff icacy was operationalized #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 558 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR TABLE 1.Descriptive Statistics for Variables in through combining measures of social cohe- the Analysis sion and informal social control. Social cohe- sion was captured using a cluster of conceptu- Variables MeanSt dev. ally related items from the PHDCN-CS mea- Self-rated health suring the respondent’s level of agreement (on Excellent .335 a f ive-point scale) with the following state- Good .443 ments: (1) “people around here are willing to Fair .175 help their neighbors,” (2) “ this is a close-knit Poor .048 neighborhood,” (3) “ people in this neighbor- Demographic background hood can be trusted, ” and (4) “ people in this Female .590 neighborhood generally don’ t get along with Age 42.74Delivered15.99 by eachIngenta other” to (reverse coded). Health-related Race informal social control was tapped through White/other .459Welch Medicalitems Library measuring (cid 82007175),the respondent’ s level of Black .349Johns Hopkinsagreement University with (cidthe following 90000553) statements: (1) Latino .192IP: 127.0.0.1“If I were sick I could count on my neighbors Income Date: 2005..04..18..17..00..to shop for groceries for me” and (2) “You can < $10,000 .091 count on adults in this neighborhood to watch $10-000–$15,000 .073 out that children are safe and don’t get in trou- $15,000–$20,000 .069 ble.” An additional informal control item asked $20,000–$25,000 .068 respondents how likely it is that neighbors $30,000–$40,000 .091 would intervene if “there was a fight in front of $40,000–$50,000 .147 your house and someone was being beaten or $50,000–$60,000 .126 threatened.” The informal control items tap $60,000–$70,000 .151 expectations for beneficial health-re lated $70,000–$90,000 .101 action as well as neighborhood supervision of >$90,000 .082 potentially hazardous conditions or violent sit- Education uations.9 The seven items were combined to 4th grade or less .014 form a single scale of health-related collective 5th to 8th grade .053 efficacy. The reliability of the collective effica- 9th to 12th grade, no diploma.106 cy scale is .73. 10 High school graduate .165 The local social support and sociability Trade or vaocational school .077 scale encompassed a number of items measur- Some college .263 ing the extent of the respondent’ s neighbor- College graduate .163 hood-based friendship networks (how many Some graduate study .033 friends the respondent has in the neighborhood Graduate degree .125 and the ratio of friends inside and outside the Married .391 neighborhood), as well as the frequency of par- Years resident in the neigh. 11.52 9.70 ties, visits, advice giving, and favor exchange Home ownership .430 among neighbors. The latter four items were Foreign born .177 constructed from questions asking respondents Health background (1) “how often do you and people in this neigh- Exercise per week borhood have parties or other get-togethers Never .336 where other people in the neighborhood are Once or twice .241 invited?,” (2) “how often do you and other peo- Two to three times .186 ple in this neighborhood visit in each other’ s Four times or more .237 homes or on the street?,” (3) “how often do you Smoking .285 and other people in the neighborhood ask each Weight problem .196 other advice about personal things such as Insurance child rearing or job openings?,” and (4) “how Private insurance .671 often do you and people in your neighborhood Medicaid .083 do favors for each other?” 11 These items are Medicare .077 intended to tap aspects of community level No insurance .169 instrumental, informational, and appraisal sup- Note: N = 3272 port as well as the level of sociability charac- #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 559 terizing the neighborhood (House, Landis, and ensure that any effects of social organization Umberson 1988). The reliability of the social variables (e.g., social support and sociability) support and sociability scale is .60. were not simply capturing the effects of a con- Subcultural orientations were captured with sistently poor neighborhood on network with- two scales—a measure of tolerance for youth- drawal and community engagement (i.e., the ful and problem behaviors and a mea- potential endogeneity of social organization). sure of . Eight items were used Prior neighborhood health was estimated from to measure tolerance for risk behavio r . the 1992–1994 waves of the MCIC by aggre- Respondents were asked their opinions regard- gating individual responses to the self-rated ing how wrong it is for nineteen-year-olds to health item to the neighborhood cluster level smoke cigarettes, use marijuana, drink alcohol, (see Appendix 1 for correlations among neigh- and get into fistfights. Though questionsDelivered were by borhoodIngenta level to variables). not asked about tolerance for such behavior by Individual level variables are taken from the adults at any age, the items neverthelessWelch cap-Medical1995, Library ’97, and (cid ’99 82007175),waves of the MCIC-MS. Of ture an underlying dimension of Johns permissive- Hopkinsthe University waves of the (cid MCIC-MS 90000553) available from ness that is likely associated with IP: neighbor- 127.0.0.1 1995 through 2000, only ’ 95, ’ 97, and ’ 99 hood level sanctioning of comparable Date: 2005..04..18..17..00..adult include the necessary individual level controls behavior. The reliability of the deviance toler- for health background. We include a variable ance scale is .61. Anomie was measured using for interview year to capture any time trends in the respondent’ s level of agreement with the the depende nt variable. Table 1 reports following statements: (1) “laws were made to descriptive statistics on the key demographic be broken,” (2) “it’s okay to do anything you background and health related items used in want as long as you don’t hurt anyone,” (3) “to the analyses. Included are measures of gender, make money , there are no right and wrong race/ethnicity (Black, Latino versus ways anymore, only easy ways and hard ways,” White/other), age, income, education level, and (4) “ fighting between friends or within marital status (married versus single or cohab- families is nobody else’s business.” This scale iting), number of years resident in the respon- is intended to capture “ anomie” (detachment dent’s current neighborhood, homeownership, from conventional norms) and fatalistic orien- and whether the respondent is foreign born. tations that may increase the likelihood of Health measures include insurance coverage health-risk behavior or failure to seek medical (Medicaid, Medicare, or no insurance coverage attention or advice (Sampson and Bartusch versus private insurer), dummy variables mea- 1998). The reliability of the anomie scale is suring a doctor-indicated weight problem and .63. current smoking, and a measure of the fre- Neighborhood disorder was included as a quency of exercise per week in four categories. control in models examining links between structural disadvantage and health in order to tap social conditions hypothesized to cue the ANALYTIC STRATEGY breakdown of social order and the potential for corresponding increases in stress levels (Ross We use Hierarchical Linear and Non-linear and Mirowsky 2001). Social and physical dis- Modeling techniques to account for the com- order could be confounded with the other plex error structure of multilevel models of mechanisms examined below (Sampson and health outcomes. The analyses reported in Raudenbush 1999). The level of disorderin the tables 2 and 3 are two-level hierarchical (ran- neighborhood was measured with four items dom intercept ) logit models of the inquiring into respondent assessments of the dichotomized self-rated health measure. 12 We prevalence of public drinking , abandone d report coefficients (in log form) in the table buildings, litter and trash, and graffiti. The but refer to percentage changes in odds in the reliability of the disorder scale is .88. The text (exponentiated coefficients). aggregation procedure for multi-item neigh- In order to correct independent neighbor- borhood survey-based measures is described in hood level measures of collective efficacy, net- more detail below (Analytic Strategy). work density , subcultural orientations, disor- Finally, we include controls for prior neigh- der, and prior neighborhood health for missing borhood health and population size. Prior data and social composition, we use empirical neighborhood healthwas included in order to bayes residuals from multilevel models of each #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 560 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR outcome (Bryk and Raudenb ush 1992; collective efficacy score ultimately included as Snijders and Bosker 1999). As an illustrative an independent variable in multilevel logit example, we describe the three level item models of the self-rated health outcome is the response model used to construct our measure standardized empirical bayes residual from the of neighborhood collective efficacy (Sampson, level three model. Empirical bayes estimates Raudenbush, and Earls 1997). At level one take into account differences in the reliability

(within individuals)— the seven items com- with which b0k has been estimated (Bryk and prising the collective eff icacy scale are mod- Raudenbush 1992). eled as follows: 6

Yijk = pjk + S apDpijk + eijk, RESULTS p = 1 Delivered by Ingenta to where Y is the response to item i of person j We begin by investigating the relationship ijk Welch Medical Library (cid 82007175), in neighborhood k, pjk is the intercept, Dpijk is between health and antecedent structural vari- a dummy variable taking on a value Johns of Hopkins 1 if ables University in multivariate (cid 90000553)context. Table 2 reports response i is to item p in the 6 itemIP: collective 127.0.0.1 the results of a series of hierarchical logit mod- Date: 2005..04..18..17..00.. efficacy scale and 0 otherwise, and eijk are els of self-rated health on individual and errors of measurement assumed to be indepen- neighborhood level characteristics. The intra- dent. Respondents who provide answers to any class correlation from the unconditional model of the scale items contribute information to the indicates that roughly 6% of the variance in analysis, though HLM weights cases with full self-rated fair or poor health (hereafter “poor” information more heavily. health) lies at the neighborhood level (appar- At level two (between individuals), respon- ently small intra-class correlations can veil rel- dent-specific latent perceptions of collective atively large effects of individual neighbor- efficacy are adjusted for individual level char- hood coefficients [Duncan and Raudenbush acteristics as follows: 1999]).13 13 Model 1 predicts self-rated health based on 2 pjk = b0k + S bqXqjk + rjk,—–rij ~ N(0,s ) demographic background factors. Consistent q = 1 with expectations and previous research, age and African-American race/ ethnicity increase where b0k is the intercept, Xqjk is the value of person-level predictor q for individual j in the likeliho od of poor self-rated health. neighborhood k, bq is the effect of q on indi- Women and foreign-born respondents also vidual j’s expected score, and rjk is an indepen- report poorer health. Income, education, and dently, normally distributed error term with interview year are strongly negatively associat- variance s2. The models adjust for thirteen ed with poor health. The association between covariates including gender, age, race/ethnicity interview year and health indicates that health (black, Latino vs. white), education, employ- improved over the five years covered by the ment status (employed vs. not employed), mar- 1995, ’ 97, and ’ 99 MCIC samples. Model 2 ital status (never married, other vs. married), adds controls for health behavior and insur- homeownership, years resident in the neigh- ance coverage. W eight problems and borhood, and number of moves in the last five Medicare, Medicaid, or no insurance coverage years. The level two model also includes a (by comparison with private insurance cover- measure of the respondent’s social embedded- age) are all positively associated with poor ness based on responses to the social support health while exercise benefits health. Though and sociability items described above. This reduced in Model 2, the effect of African- control ensures that the neighborhood level American race/ethnicity remains after control- measure of collective efficacy is not confound- ling for health background, indicating that the ed with individual level social support and odds of reporting poor health are 47% higher sociability. Finally, adjusted neighbor hood than the odds for Whites. 14 intercepts are modeled as follows: Model 3 includes only neighborhood struc- tural characteristics in order to examine the b0k = g00 + u0k,–––u0k ~ N(0,t00) effect of these variables without individual level controls. Controlling only for interview where g00 is the grand mean and u0k is a level year, poverty , residential stability, and immi- three random effect . The neighborhood level grant concentration are all positively associat- #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 561

TABLE 2.Hierarchical Logit Models of Self-Rated Fair or Poor Health on Individual Background, Health Behavior,Insurance Coverage,and Neighborhood Characteristics Fair or Poor Health Independent Variables .01 .02 .03 .04 .05 .06 Individual level Female .216* .146 .— .145 .144 .147 (.097) (.100) (.099) (.101) (.103) Age .028**.025** .— .025**.026** .026** (.004) (.005) (.005) (.005) (.005) Black .468**.384** .— .425** .238 .250 (.104) (.108) (.134) (.142) (.142) Latino .172 .136 .— .140 .086 .072 (.153)Delivered(.160) by Ingenta to (.159) (.157) (.158) Income –.076**Welch– .051*Medical Library.— (cid–.054* 82007175),– .046*– .047* (.021) (.022) (.021) (.022) (.022) Education –.245**Johns –Hopkins .216** University.— –(cid.215** 90000553)– .200*– .199** (.030)IP: 127.0.0.1(.030) (.032) (.031) (.032) Married .150Date: 2005..04..18..17..00...145 .— .132 .137 .140 (.104) (.105) (.105) (.105) (.107) Years in the neighborhood –.003 –.003 .— –.004 –.005 –.005 (.008) (.008) (.008) (.008) (.008) Home ownership –.118 –.058 .— –.147 –.133 –.142 (.115) (.117) (.123) (.124) (.124) Foreign born .369**.467** .— .482**.493** .497** (.132) (.136) (.136) (.134) (.136) Interview year –.169**– .187**– .189**– .188**– .188**– .189** (.032) (.003) (.029) (.033) (.033) (.034) Exercise frequency per week .— –.095* .— –.095*– .096*– .098* (.039) (.039) (.039) (.040) Smoking .— .197 .— .212 .192 .197 (.106) (.106) (.107) (.108) Weight problem .— .756** .— .752**.740** .746** (.107) (.106) (.107) (.107) Medicare .— .558** .— .572**.549** .554** (.168) (.170) (.170) (.171) Medicaid .— .470** .— .476**.463** .461** (.165) (.166) (.169) (.170) No insurance .— .442** .— .444**.440** .444** (.135) (.132) (.134) (.135) Neighborhood level Total population .— .— .018 .022 .039**.032* (.013) (.014) (.014) (.014) Residential stability .— .— .288**.107* .183**.249** (.046) (.052) (.054) (.062) Immigrant concentration .— .— .139** .014 –.011 .007 (.052) (.079) (.079) (.077) Poverty .— .— .019**– .003 .— – (.003) (.003) Affluence .— .— .— .— –.014*– .014* (.006) (.005) Stability*Affluence .— .— .— .— .— –.010* (.004) Intercept –1.513**– 1.557**– 1.243**– 1.576**– 1.549**– 1.556** (.052) (.054) (.048) (.059) (.059) (.057) Intercept variance component.031 .057 .099* .039 .050 .026 * p < .05; ** p < .01 (two-tailed tests). Standard errors in parentheses. Note: Neighborhood N = 339; Individual N = 3,272. ed with poor health. While the effects of pover- Model 4 includes the three structural factors ty and immigrant concentration are consistent as well as demographic and health background with the basic social disorganization perspec- characteristics. In contrast to a number of tive, the negative effect of residential stability recent studies (Robert 1999), model 2 indi- is opposite of the expected positive effect. cates that the effect of poverty on health in #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 562 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR model 1 is due to the social composition of Model 6 tests Wilson’s expectation that neigh- Chicago neighborhoods. 15 That is, according to borhood economic structure (represented here these data, neighborhood poverty exerts no by the significant effect of affluence 16) condi- independent effect on self-rated health when tions the effect of residential stability on individual level characteristics are controlled. health. Indeed, the interaction between afflu- Residential stabilit y, on the other hand, ence and stability is significant, indicating that remains a signif icant negative predictor of as affluence increases, the positive effect of health after individual controls have been stability on poor health declines. Alternatively, included. Individual level covariates remain as stability increases, the effect of affluence on largely unchanged in Model 4, though the poor health is increasingly negative. 17 effect of individual level African-American The conditional impact of residential stabil- race/ethnicity increases somewhatDelivered from Model by ityIngenta at varying to levels of neighborhood affluence 2. Welch Medicalis presented Library in(cid Figure 82007175), 1.The predicted proba- Model 5 replaces poverty with a measure of Johns Hopkinsbility University of fair/ poor (cid health 90000553) is plotted from –1.5 the prevalence of affluence at the neighbor- (low) to 1.5 (high) standard deviations on the hood level. The coefficient for affluenceIP: 127.0.0.1 is sig- residential stability scale for three levels of nificant and negative, indicating Date:that increas- 2005..04..18..17..00.. community affluence (corresponding to the ing neighborhood affluence is health enhanc- mean and 1.5 standard deviations above and ing, independent of individual demographic and health background. Including affluence below the mean). Consistent with Wilson’ s also results in a significant positive effect of interactive model, high levels of residential population size on poor health and increases stability are positively associated with poor the magnitude of the residential stability effect. health at low levels of community affluence Notably, inclusion of neighborhood affluence (2%): The predicted probability of poor health reduces the magnitude of the coefficient for increases significantly from .13 to .22 with a African-American race (by comparison with movement from 1.5 standard deviations below Model 4) by 44% and renders it insignificant the mean on the residential stability scale to at the conventional level. 1.5 standard deviations above the mean. By Finally, the positive effect of residential sta- contrast, the significant positive effect of resi- bility on poor health suggests that stable neigh- dential stability on poor health declines in borhoods are not health-protective, as social communities with 18% affluent residents and disorganization theory would hypothesi ze. becomes negative (though nonsignificant [ p >

FIGURE 1.Probability of Fair/Poor Health by Neighborhood Residential Stability and Affluence Level #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 563 .05]) in communities with 34% affluent resi- vidual health, above and beyond individual dents. demographic and health background. Thus, we find that neighborhood affluence is a more powerful predictor of self-rated health than poverty, but that its effect is condi- DISCUSSION tioned by residential stability. Moreover, con- trolling for neighborhood affluence substan- Considerable research interest in the impact tially diminishes the magnitude of the residual of neighborhood characteristics on health out- association between African-Am erican comes has generated a number of empirical race/ethnicity and self-rated health. efforts to assess these effects. This research has We next consider the role of intervening largely focused on the effects of prevalent social process mechanisms in explaining Delivered the by neighborhoodIngenta to economic deprivation, however, effects of structural characteristics on health. to the exclusion of other potentially relevant Table 3 reports the results of tests Welchof the social Medicalstructural Library characteristics. (cid 82007175), Moreover, to date, organizational and cultural factors Johns hypothe- Hopkinsresearch University on neighborhoods (cid 90000553) and health has not sized to mediate neighbo rhoodIP: structur 127.0.0.1al assessed the social mechanisms hypothesized effects. Models 1 through 3 add Date:measures 2005..04..18..17..00.. of to channel the impact of neighborhood struc- neighborhood social support and sociability, ture on health outcomes. W e extend existing risk behavior tolerance and anomie, and col- research on the community context of health lective eff icacy in sequence. In addition to by examining multiple dimensions of neigh- measures of neighborhood structure, controls borhood structur e—includin g concentra ted for neighborhood disorder, population size, affluence, residential stability, and ethnic het- and prior neighborhood health are also includ- erogeneity—in addition to poverty and exam- ed.18 As is clear, none of the intervening social ine a number of social processes hypothesized process measures is significantly associated to be relevant in linking neighborhood struc- with health, with the sole exception of collec- ture with health. The latter include local social tive efficacy. Model 3 indicates that collective support and sociability, the emergence of sub- efficacy is negatively, and nontrivially associ- cultures of tolerance for deviant risk behavior ated with poor self-rated health. A two stan- and anomie, and collective efficacy. Research dard deviation increase on the collective effi- to date has not examined these distinct mecha- cacy scale results in a 27% reduction in the nisms simultaneously in order to isolate the odds of reporting poor self-rated health. The community level social factors that are rele- introduction of the collective eff icacy mea- vant for health. sure, however, results in negligible changes in Our analyses brought several data resources the coefficients for neighborhood structural together in order to investigate neighborhood characteristics, suggesting that the effects of effects on self-rated health. Neighborhood affluence and residential stability do not work level indicators of structural disadvanta ge principally through collective efficacy. from the 1990 census were combined with Finally, model 4 removes health background high quality indicators of social and cultural and insurance coverage from the individual processes derived from the PHDCN-CS. These level model while retaining the full battery of measures were merged with individual level neighborhood effects. This strategy allows for data and a measure of prior neighborhood an assessment of the possible extent to which health taken from the MCIC-MS. neighborhood effects work through health Our findings address a number of issues rel- behaviors to affect health outcomes. Model 4 evant to the discussion surrounding communi- offers little evidence that neighborhood effects ty effects on health. First, our analyses indicate are influencing self-rated health through health that the effect of concentrated poverty of behaviors, with the possible exception of the health is due to the social composition of poor coefficient for anomie. The effect of anomie neighborhoods. In the context of a model achieves marginal signif icance ( p < .10) in excluding individual background factors, a Model 4, suggesting that health behaviors may measure of neighborhood poverty was highly partially mediate this effect. In sum, the analy- significant and positively associated with poor ses presented in Tables 2 and 3 offer evidence self-rated health. Yet, when we controlled for that neighborhood affluence, stability, and col- individual level characteristics, the impact of lective eff icacy contribute, uniquely , to indi- poverty on health was rendered insignificant. #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 564 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR

TABLE 3.Hierarchical Logit Models of Self-Rated Fair or Poor Health on Individual Background, Health Behavior,Insurance Coverage,and Neighborhood Characteristics Independent Fair or Poor Health

Variables .01 .02 .03 .04 Individual level Female .142 .143 .143 .209* (.101) (.100) (.101) (.098) Age .026** .026** .026** .030** (.005) (.005) (.005) (.005) Black .261 .256 .281 .382** (.146) (.147) (.145) (.137) Latino .070 .064 .057 .117 (.165) (.158) (.164) (.150) Income Delivered–.046* by Ingenta–.046* to –.046* –.069* Welch(.022) Medical Library(.021) (cid 82007175),(.022) (.021) Education Johns–.199** Hopkins University–.197** (cid 90000553)–.199** –.231** (.032) (.031) (.032) (.030) Married IP:.154 127.0.0.1 .156 .159 .123 Date:(.106) 2005..04..18..17..00..(.105) (.106) (.104) Years in the neighborhood –.005 –.005 –.005 –.005 (.008) (.007) (.008) (.008) Own home –.148 –.144 –.152 –.199 (.123) (.121) (.122) (.120) Foreign born .511** .509** .512** .394** (.140) (.136) (.138) (.131) Interview year –.191** –.191** –.187** –.149** (.033) (.033) (.033) (.028) Exercise frequency –.098 –.095 –.097* .— (.039) (.039) (.040) Smoking .201 .201 .203 .— (.106) (.105) (.107) Weight problem .746** .743** .736** .— (.107) (.107) (.106) Medicare .562** .561** .561** .— (.173) (.171) (.175) Medicaid .457** .449** .455** .— (.166) (.168) (.167) No insurance .439** .448** .440** .— (.133) (.130) (.132) Neighborhood level Total population .033* .032* .030* .033* (.014) (.014) (.014) (.014) Residential stability .250** .236** .244** .234** (.061) (.060) (.061) (.060) Immigrant concentration .007 –.008 –.017 .036 (.078) (.081) (.080) (.067) Affluence –.016* –.015* –.015* –.019** (.007) (.007) (.006) (.006) Stability*Affluence –.010** –.010** –.010** –.008* (.004) (.004) (.004) (.004) Prior neighborhood health .029 .020 .028 –.005 (.046) (.046) (.046) (.042) Disorder –.018 –.019 –.091 –.111 (.075) (.075) (.080) (.079) Networks/exchange .020 .017 .097 .112 (.058) (.058) (.066) (.067) Tolerance of risk behavior .— –.026 –.032 –.022 (.046) (.047) (.046) Anomie .— .072 .079 .100 (.050) (.050) (.052) Collective efficacy .— .— –.155* .136* (.071) (.063) Intercept –1.552** –1.552** –1.558** –1.453** (.058) (.058) (.058) (.054) Intercept variance component .027 .024 .003 .003 * p < .05; ** p < .01 (two-tailed tests). Standard errors in parentheses. Note: Neighborhood N = 339; Individual N = 3,272. #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 565 This finding is at odds with a number of recent dimensions of neighborhood economic struc- studies that demonstrate the unique impact of ture that may be relevant for health outcomes. poverty on health after controlling for individ- Fourth, tests of the mechanisms hypothe- ual factors (Kawachi, Kennedy , Lochner, and sized to link neighborhoo d structure with Prothrow-Stith 1997; Malmstrom, Sundquist, health resulted in a significant effect only for a and Johansson 1999; Robert 1999). measure of collective efficacy. Though a num- Second, in contrast to the effect of poverty, ber of researchers have hypothesized that ben- we find that a measure of neighborhood afflu- eficial social ties at the neighborhood level ence exerts unique protective effects on health, will have consequences for community health, independent of individual factors and neigh- we found no direct effect of a measure of borhood structural controls. The effect of afflu- social support and sociability on self-rated ence, by comparison with poverty Delivered, indicates by health.Ingenta T ests to of the impact of subculture on that the prevalence of middle and upperWelch middle Medicalhealth Library also yielded (cid 82007175), no unique effects of neigh- class residents in urban communities Johns isHopkins an borhood University level measures (cid 90000553) of anomie and toler- important structural factor influencingIP: 127.0.0.1 health- ance for deviant behavior. This might be promoting conditions. These dataDate: suggest 2005..04..18..17..00.. that expected to the extent that we control for the measures of the proportion of residents who effects of individual level health-risk behaviors are poor in a neighborhood may be capturing, on self-rated health— a key individual level in part, the influence of the absence of resi- mechanism through which the effects of neigh- dents with resources. Consistent with Wilson’s borhood health-related subcultures are hypoth- emphasis on the potentially detrimental effects esized to flow. Even before including individ- of constrained mobility in disadvantaged urban ual controls for health risk behavior, however, neighborhoods, affluence conditioned the subcultural measures were not signif icantly effect of residential stability. In neighborhoods predictive of self-rated health— only anomie with low levels of affluence, stability increases achieved marginal significance in the absence the likelihood of poor health. This effect of health background controls. Although these declines, however, as affluence increases. This analyses offer little support for hypotheses finding is consistent with other recent research regarding subcultural influences on health, demonstrating the conditional effects of resi- more effective measurement of the situational dential stability on health outcomes (Ross, and group basis of alternative cultural orienta- Reynolds, and Geis 2000) and calls attention to tions (as opposed to aggregated survey-based the detrimental consequences of combining assessments of individual attitudes) may yield sustained immobility with the absence of mid- different findings. dle-class residents. The latter may have the The impact of collective eff icacy may be social and economic resources with which to channeled through a number of mechanisms. sustain local institutions, benefiting the neigh- First, collective efficacy may orient neighbor- borhood as a whole. hood residents to the well-being of other mem- Third, the association between individual bers of the community. As Putnam (2000) has level African-Am erican race/ ethnicity and noted, strong attachme nts to communi ty poor health was substantially reduced when increase the likelihood that indicators of health affluence was controlled, though unaffected by problems may receive notice (e.g., a communi- inclusion of our measure of neighborhood ty member who is missing from church or a poverty. Consistent with previous research local meeting is more likely to elicit the con- exploring the impact of affluence on race dif- cern of neighbors). Moreover, sentiments of ferences in health among the elderly (Cagney, solidarity and trust may lead community mem- Browning, and Wen 2002), this finding sug- bers to act on others behalf independently of a gests that models linking individual measures prior network tie. Second, the psychosocial of disadvantage with health outcomes must benefits of trust may reduce stress and facili- simultaneously consider the prevalence of tate use of local space for recreational and advantage in the environment with which indi- exercise purposes. Indeed, some have found vidual level factors may be confounded. A sim- that the perception of social support (more ple assumption that race is confounded with directly related to the concept of trust) is more the prevalence of poverty ignores the multiple strongly associated with health than actual #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 566 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR social support (Dunkel-Schetter and Bennett from the MCIC-MS data. Moreover, the col- 1990). lective efficacy measure does not capture actu- Third, as some have argued, collective effi- al instances of action on behalf of community cacy may signal a community’ s capacity to goals, but rather the perception that such mobilize on behalf of a wide range of impor- action would be taken under certain condi- tant neighbo rhood goals (Sampson, tions. Finally, we consider the effects of neigh- Raudenbush, and Earls 1997). This capacity borhood characteristics on a single indicator of may extend to the provision of high quality self-rated health status; the effects of neigh- health care and health-re levant services. borhood structure and social process measures Mobilization capacity may also foster access may differ by the specific health outcome con- to resources outside the neighborhood that aid sidered. in correcting local physical hazards such as Future research will hopefully clarify a infrastructural decay (abandoned Delivered buildings, by Ingenta to Welch Medicalnumber Library of outstanding (cid 82007175), questions accompany- crumbling sidewalks, inadequate traffic sig- ing or motivated by these analyses. First, the nage, etc.). The quality and availability Johns Hopkins of University (cid 90000553) IP: 127.0.0.1 impact of objective measures of medical infra- health services and extra-community resources structure on health may contribute to the may be a critical pathway through Date: which 2005..04..18..17..00..explanatory power of models that include neighborhood collective efficacy operates. neighborhood effects. The capacity to attract Indeed, this pathway may also account, in part, and support local medical facilities, as noted, for the robust effect of neighborhood affluence may be associated with both affluence and col- on health. lective efficacy and potentially mediate some The study was not without limitations. First, proportion of their effects on individual health. our sample is restricted to the city of Chicago. While the data provide an unpreceden ted Second, employing longitudinal data on health opportunity to consider the influence of both outcomes will provide an opportunity both to neighborhood structural and social process establish the causal influence of neighborhood measures on health, our capacity to generalize factors with more certainty and to tease out the from the analyses is limited. Multiple city or indirect pathways—both individual and neigh- national samples would also likely capture borhood level—through which neighborhoods greater variation across neighborhoods in the operate to affect health. Third, national data on health of residents. Second, our individual neighborhood characte ristics and multiple level data are cross-sectional. Despite relative- health outcomes will allow for generalizable ly extensive health controls, we are still unable conclusions and more precise (and simultane- to rule out selection as a confounding process ous) estimation of the unique effects of dimen- in the association between neighborhood fac- sions of neighborhood structure (e.g., poverty tors and health. Third, though our measure of and affluence). These findings and the results collective efficacy is purged of its association of future analyses will add to our accumulating with individual level social support and socia- knowledge regarding the significance of macro bility, we were not able to include a compara- structural characteristics and social processes ble individual level control derived directly for the spatial organization of health.

Appendix. Correlations Among Neighborhood Level Variables in the Analysis (N=339) Variables .01 .02 .03 .04 .05 .06 .07 .08 .09 .10 .11 .13 01. Total population 1.000 02. Residential stability –.153 1.000 03. Immigrant concentration– .040– .2121.000 04. Poverty –.229– .303– .1111.000 05. Affluence .276.366 – .120– .7681.000 06. Stability*Affluence –.179.463 – .008– .057.190 1.000 07. Neighborhood health (’92–’94) – .094– .020– .037.055 – .136.047 1.000 08. Networks/exchange –.028.169 .021 – .024.302 .176 – .0081.000 09. Tolerance of risk behavior– .004– .181– .067.076 – .072– .130– .118.078 1.000 10. Subcultural cynicism –.018.033 .105 .093 – .181– .019.026 .034 .055 1.000 11. Disorder –.199– .290.114 .773 – .766– .142.058 – .243.107 .148 1.000 12. Collective efficacy .019.309 – .116– .418.537 .202 .038 .599 – .079– .025– .5831.000 #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 567 NOTES unique effects difficult to assess in multi- variate models. This problem has led some 1.Like the original social disorganization per- to argue for the use of factor analytic spective, the collective efficacy approach approaches—a strategy we employ for the was conceived to explain variations in measurement of residential stability and neighborhood crime rates. Therefore, col- ethnic heterogeneity. Because we are con- lective efficacy with respect to the control cerned with the comparative effects of of neighborhood crime should not be seen poverty and affluence, however, we chose as necessarily equivalent to the community to focus on single item measures of eco- capacity to promote health. At the same nomic status in order to avoid complicat- time, evidence suggests that collective effi- ing interpretation of their relative effects cacy is a generalizable resourceDelivered capable of by Ingentaon health. to influencing a wide range of outcomes 07.The analysis employed alpha-scoring fac- including self-repor ted physical Welch health Medical Librarytor analysis (cid 82007175),with an oblique rotation. (Browning and Cagney 2002) andJohns violence Hopkins UniversityScores from (cid principal 90000553) components analy- between intimate partners IP:(Browning 127.0.0.1 ses yielded the same pattern of effects in 2002). Date: 2005..04..18..17..00..multivariate analyses of health. 2.Respondents were given the following defi- 08.Eigenvalues for the two factors exceeded nition of “ neighborhood:” “ By neighbor- 1. hood . . . we mean the area around where 09.Though our measure of collective efficacy you live and around your house. It may does not directly tap instances of pro- include places you shop, religious or public social action on behalf of the health of institutions, or a local business district. It is local residents, this approach to measuring the general area around your house where collective efficacy is likely to be highly you might perform routine tasks, such as correlated with health problems that spur shopping, going to the park, or visiting with action. To the extent that these precipitat- neighbors.” ing health problems are differentially dis- 3.The response rate for the PHDCN-CS was tributed across neighbo rhoods, actual 75%. mobilization to address them may be con- 4.The response rate for the MCIC-MS aver- founded with poor health. One approach aged around 55% across the 10 cross-sec- to measuring action on behalf of commu- tional samples. Because the MCIC-MS did nity health would be to measure neighbor- not achieve as high a response rate as the hood responses to exogenou s health PHDCN-CS, we compared the latter with shocks to which all communit ies are combined 1993, ’ 94, ’ 95, and ’ 96 MCIC- exposed (Klinenberg 2002). MS samples. The distributions across basic 10.Distinct from individual level reliability demographic characteristics (gender, age, (e.g, Cronbach’ s alpha), the reliability of and race) were quite similar. neighborhood level scale scores is depen- 5.W e also investigated the possibility that the dent upon both the sample size within the relationship between self-rated health and neighborhood as well as the proportion of other health outcomes varied by neighbor- the total variance that is between (vs. with- hood (calling into question efforts to in) neighborhoods. The reliability esti- explain variation in self-rated health across mates of neighborhood scales are compa- context). A series of random slope models rable to reliabiliti es reported in other examining potential variation in the rela- recent neighbo rhood research (see tionship between self-rated health and self- Sampson, Morenoff, and Earls 1999). reports of having ever been told by a doctor 11.Response categories for the latter four that the respondent has asthma, a heart con- items were “never, rarely , sometimes, or dition, or high cholesterol offered no evi- often.” dence of significant variability in slope 12.W e investigated whether the effects of coefficients across neighborhoods. individual level covariates varied across 6.The well known problem of collinearity neighborhoods by examining a series of among macro level indicators of socioeco- random slope models. T ests of signifi- nomic status, including income, education, cance on variance components for random and occupati onal status, renders their slopes offered no evidence of significant #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 568 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR variability in the effects of individual level REFERENCES covariates across neighborhoods. 13.While we report the intercept variance Andersen, Ronald M., Aday,L.A., L yttle, C.S., across models, these estimates should be Cornelius, L.J., & Chen, M.-S. 1987. interpreted with caution. The increase in Ambulatory Care and Insurance Coverage in an Era of Constraint . Chicago: Pluribus Press. the variance component with the addition Anderson, Elijah. 1990. Streetwise. Chicago: of neighbo rhood level predicto rs, for University of Chicago Press. instance, implies an impossibility— i.e., Anselin, Luc. 1988. Spatial Econometrics: Methods that the percentage of variance explained and Models. Dordrecht , Germany: Kluwer has actually decreased across more fully Academic. specified models. Volatility in the estimate Benyamini, Y. and E. L. Idler. 1999. “ Community of the variance component, however, Delivered can by IngentaStudies toReporting Asssociation Between Self- Rated Health and Mortality.” Research on Aging occur when the residual variance Welch at Medical the 21:477–Library500. (cid 82007175), neighborhood level is low (Snijders Johns Hopkins and Berkman, University Lisa F (cid. 1995. 90000553) “ The Role of Social Boskers). Rather than rely IP:on intercept 127.0.0.1 Relations in Health Promotion.” Psychosom Med variance estimates and associatedDate: signifi- 2005..04..18..17..00..57:245–54. cance tests, we focus on more powerful Berkman, Lisa F. and L. Breslow. 1983. Health and tests of theoretically relevant coefficients. Ways of Living . New Y ork: Oxford University 14.The spatial clustering of self-reported Press. health at the neighborhood level led us to Berkman, Lisa F., Thomas Glass, Ian Brissette, and Teresa E. Seeman. 2000. “ From Social investigate the possibil ity that spatial Integration to Health: Durkheim in the New dependency was operating in the data. A Millennium.” Social Science & Medicine Moran’s I test for spatial correlation in the 51:843–57. residuals (Anselin 1988) from Model 3 of Berkman, Lisa F . and S. L. Syme. 1979. “ Social Table 3, however, was not significant (p > Networks, Host Resistance and Mortality: A .05). Dummy variables indicating region Nine Year Follow-up Study of Alameda County of the City (north, west, and south-corre- Residents.” American Journal of Epidemiology 109:186–204. sponding to major ethnic/ racial divisions Blair, Steven N., Harold W . Kohl III, Ralph S. within the city) were also not significant Paffenbarger, Debra G. Clark, Kenneth H. predictors of self-rated health. Cooper, and Larry W. Gibbons. 1989. “Physical 15.Entering demographic characteristics by Fitness and All-Cause Mortality.” Journal of the stages indicates that race/ ethnicity and American Medical Association 262:2395–2401. education account for the associati on Browning, C. R. 2002. “ The Span of Collective between neighborhood poverty and self- Efficiency: Extending Social Disorganization Theory to Partner Violence.” Journal of rated health. Marriage and the Family 64(4):833–50. 16.W e examined the interacti on between Browning, C. R. and K. Cagney. 2002. “Collective poverty and stability separately. The coef- efficacy and health: neighborhood ficient was not significant. and self-rated physical functioning in an urban 17.Substituting, as measures of neighborhood setting.” Journal of Health and Social Behaviors affluence, the percentage college educated 43(4):383–99. in the neighborhood and (separately) a Bryk, Anthony S. and Stephen W . Raudenbush. composite index of education and income 1992. Hierarchical Linear Models: Applications and Data Analysis Methods . Newbury Park, N.J.: yield the same pattern of interaction with Sage. residential stability in comparable multi- Bursik, Robert J. 1988. “Social Disorganization and variate analyses of fair or poor self-rated Theories of Crime and Delinquency: Problems health. and Prospects.” 26:519–51. 18.Though we include the prior neighbor- Cagney, K., C. R. Browning, and M. W en. 2002. hood health measure to account, in part, “Race and self-rated health at older ages: what for any possible endogeneity of neighbor- difference does the neighbor hood make?” Unpublished manuscript . hood characterist ics, models with and Davey Smith, G., C. Hart, G. Watt, D. Hole, and V. without this control result in negligible Hawthome. 1998. “Individual social class, area- changes in coefficient magnitudes and sig- based deprivation, cardiovscular disease risk fac- nificance levels. tors, and mortality: the Renfrew and Paisley #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 569

study.” Journal of Epidemiology Community tions in health outcomes: artefact or ecology.” Health 52:399–405. International Journal of Epidemiolo gy Diez-Roux, A. V ., F . J. Nieto, C. Muntaner, H. A. 20:251–58. Tyroler, G. W . Comstock, E. Shahar, L. S. Idler, E. L. and R. J. Angel. 1990. “Self-rated Health Cooper, R. L. Watson, and M. Szklo. 1997. and Mortality in the NHANES-1 Epidemiologic “Neighborhood environme nts and coronary Follow-up Study .” American Journal of Public heart disease: a multilevel analysis.” American Health 80:446–52. Journal of Epidemiology 146:48–63. Idler, E. L. and Y. Benyamini. 1997. “ Self-Rated Duncan, Greg J. and Stephen Raudenbush. 1999. Health and Mortality: A Review of T wenty- “Assessing the Effects of Context in Studies of Seven Community Studies.” Journal of Health Child and Y outh Development.” Educational and Social Behavior 38:21–37. Psychologist 34:29–41. Idler, E. L. and S. Kasl. 1995. “ Self-Ratings of Dunkel-Schetter, C. and T . L. Bennett. 1990. Health: Do They Also Predict Change in “Differentiating the cognitive and Delivered behavioral by IngentaFunctional to Ability.” Journal of Gerontology: aspects of social support.” Pp. 267–Welch96 in Social Medical SocialLibrary Sciences (cid50B:S344– 82007175),S353. Support: An Interactional View , editedJohns by HopkinsB. R. Johnson, University RJ and FD (cid Wolinsky 90000553). 1994. “Gender, race, Sarason, I. G. Sarason, and G. R. IP: Pierce. 127.0.0.1 New and health: The structure of health status among York: John Wiley. Date: 2005..04..18..17..00..older adults.” The Gerontologist 34:24–35. Durkheim, Emile. 1979. . New Y ork, NY : Jones, K. and C. Duncan. 1995. “ Individuals and The . their ecologies: analyzing the geography of Ecob, R. 1996. “A multilevel modeling approach to chronic illness within a multilevel modelling examining the effects of area of residence on framework.” Health & Place 1:27–40. health and function ing.” J R Stat Soc A Kaplan, G, V . Barell, and A. Lusky. 1988. 159:61–75. “Subjective State of Health and Survival in Ferraro, K. F., M. M. Farmer, and J. A. Wybraniec. Elderly Adults.” Journal of Gerontology: Social 1997. “ Health Trajectori es: Long-T erm Sciences 43:S114–120. Dynamics Among Black and White Adults. ” Kaplan, G, W. J. Strawbridge, T. Camacho, and R. Journal of Health and Social Behavior Cohen. 1993. “ Factors Associated with Change 38:38–54. in Physical Functioning in the Elderly: A Six- Finch, B.K., RA Hummer, M Reindel, and W A Year Prospective Study .” Journal of Aging and Vega. 2002. “Validity of self-rated health among Health 5:40–53. Latino(a)s.” American Journal of Epidemiology Kasarda, J. D. and M. Janowitz. 1979. “Community 155:755–59. Attachment in Mass .” American Fitzpatrick, K. and M. Lagory . 2000. Unhealthy Places: the Ecology of Risk in the Urban Sociological Review 39:329–39. Landscape. New York: Routledge. Katz, Lawrence F., Jeffrey R. Kling, and Jeffrey B. Gibson, RC. 1991. “ Race and the self-reported Liebman. 2001. “ Moving to opportuni ty in health of elderly persons. ” Journal of gerontol- Boston: Early results of a randomized mobility ogy: Social Sciences 46:S235–42. experiment.” Quarterly Journal of Economics Goldstein, MS, JM Siegel, and R Boyer. 1984. 116:607–54. “Predicting Changes in Perceived Self-Reported Kawachi, I and Lisa F . Berkman. 2000. “ Social Health Status.” American Journal of Public Cohesion, Social Capital, and Health.” in Social Health 74:611–14. Epidemiology , edited by L. F . Berkman and I. Haan, M., G.A. Kaplan, and T . Camacho. 1987. Kawachi. Oxford: Oxford University Press. “Poverty and health: prospective evidence from Kawachi, I, B.P . Kennedy , K. Lochner, and D . the Alameda County Study. ” American Journal Prothrow-Stith. 1997. “ Social Capital, Income of Epidemiology 125:989–98. Inequality, and Mortality.” American Journal of Hawe, Penelope and Alan Shiell. 2000. “ Social Public Health 87:1491–98. Capital and Health Promotion: A Review.” Klinenberg, Eric. 2002. Heatwave: A Social Autopsy Social Science & Medicine 51:871–85. of Disaster in Chicago . Chicago: The University House, J.S., K.R. Landis, and D. Umberson. 1988. of Chicago Press. “Social Relations hips and Health.” Science Kornhauser, Ruth Rosner. 1978. Social Sources of 241:540–45. Delinquency . Chicago: University of Chicago House, J.S., C. Robbins, and H.L. Metzner. 1982. Press. “The Association of Social Relationshipsand Krieger, N. 1992. “ Overcoming the absence of Activities with Mortality: Prospective Evidence socioeconomic data in medical records: valida- from the Tecumseh Community Health Study. ” tion and application of a census-based method- American Journal of Epidemiology 116:123–40. ology.” American Journal of Public Health House, James. 1981. Work Stress and Social 82:703–10. Support. Reading, MA: Addison-Wesley. LeClere, Felicia B., Richard G. Rogers, and Humphreys, K. and R. Carr-Hill. 1991. “Area varia- Kimberley Peters. 1997. “Ethnicity and mortali- #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning 570 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR

ty in the United States: individual and communi- Sampson, Robert J. 1997. “ The Embeddedness of ty correlates.” Social Forces 76:169–367. Child and Adolescen t Developmen t: A ———. 1998. “ Neighborhood Social Context and Community-Level Perspective on Urban Racial Differences in W omen’s Heart Disease Violence.” Pp. 31–77 in Violence and Childhood Mortality.” Journal of Health and Social in the Inner City , edited by J. McCord. Behavior 39:91–107. Cambridge: Cambridge University Press. Leventhal, T ama and Jeanne Brooks-Gunn. 2001. Sampson, Robert J. and Dawn Jeglum Bartusch. “Moving to Opportunity: An experimental study 1998. “ Legal Cynicism and (Subcult ural?) of neighborho od effects on mental health.” Tolerance of Deviance: The Neighborh ood Unpublished manuscript . Context of Racial Differences. ” Law & Society Malmstrom, Marianne, Jan Sundquist, and Sven- Review 32:777–804. Erik Johansso n. 1999. “ Neighborhood Sampson, Robert J., Jeffrey D. Morenoff, and Felton Environment and Self-Reported Health Status: A Earls. 1999. “ Beyond Social Capital: Spatial Multilevel Analysis.” AmericanDelivered Journal of by Ingenta to Dynamics of Collective Eff icacy.” American Public Health 89:1181–86. Welch Medical Library (cid 82007175), Sociological Review 64:633–60. Mossey, J.M., & Shapiro, E. (1982). Self-rated Johns HopkinsSampson, University Robert (cidJ. and 90000553) Stephen W . Raudenbush. health: a predictor of mortality amongIP: 127.0.0.1the elder- ly. American Jounal of Public Health , 72, 1999. “Systematic Social Observation of Public 800–08. Date: 2005..04..18..17..00..Spaces: A New Look at Disorder in Urban Orth-Gomer, K. and J. Johnson. 1987. “ Social Neighborhoods.” American Journal of Sociology Network Interaction ande Mortality: A Six Year 105:603–52. Follow-up of a Random Sample of the Swedish Sampson, Robert J., Stephen W . Raudenbush, and Population.” J Chronic Dis 40:949–57. Felton Earls. 1997. “Neighborhoods and Violent Paffenbarger, Ralph S., Robert T. Hyde, Alvin Wing, Crime: A Multilevel Study of Collective I-Min Lee, Dexter L. Jung, and James Kampert. Efficacy.” Science 227:918–23. 1993. “The Association of Chagnes in Physical- Shaw, Clifford R. and Henry D. McKay . 1969. Activity Level and Other Lifestyle Juvenile Delinque ncy and Urban Areas . Characteristics with Mortality Among Men.” Chicago: The University of Chicago Press. New England Journal of Medicine 329, 538–45. Sloggett, A. and H. Joshi. 1994. “Higher mortality Putnam, Robert. 2000. Bowling Alone: The Colapse in deprived areas: community or personal disad- and Revival of American Community . New York: vantage?” BMJ 309:1470–74. Simon & Schuster. Snijders, T om and Roel Bosker. 1999. Multilevel Raudenbush, Stephen W . and Anthony S. Bryk. Analysis: An Introduction to Basic and Advanced 2002. Hierarchical Linear Models: Applications Multilevel Modeling . London: Sage. and Data Analysis Methods . Thousand Oaks, NJ: Waitzman, N.J. and K.R. Smith. 1998. “Phantom of Sage. the area: poverty-area residence and mortality in Robert, Stephanie A. 1998. “ Community-level the United States.” American Journal of Public socioeconomic status effects on adult health.” Health 88:973–76. Journal of Health and Social Behavior Wilson, IB and S Kaplan. 1995. “Clinical Practice 39:18–37. and Patient’s Self-Reported Health Status: How ———. 1999. “ Socioeconomic Position and are the T wo Related?” Medical Care Health: The Independ ent Contribu tion of 33(suppl):S209–S214. Community Socioeconomic Context. ” Annual Wilson, William J. 1987. The Truly Disadvantaged: Review of Sociology 25:489–516. The Inner City, the Underclass, and Public Rook, K.S. 1990. “Social Relationships as a Source Policy. Chicago: University of Chicago Press. of Companion ship: Implicati ons for Older Adults’Psychological.” Pp. 221– 50 in Social ———. 1996. When Work Disappears . New York: Support: An Interactional View , edited by B. R. Alfred A. Knopf. Sarason, T. G. Sarason, and G. R. Pierce. New Wolinsky, Fredric D, and William M. Tierney. 1998. York: John Wiley. “Self-rated health and adverse health outcomes: Ross, C.E. and J. Mirowsky. 2001. “Neighborhood An exploration and refinement of the trajectory Disadvantage, Disorder, and Health.” Journal of hypothesis.” Journal of Gerontology: Social Health and Social Behavior 42:258–76. Sciences 53B(6): S336–S340. Ross, C.E., J. R. Reynolds, and K. J. Geis. 2000. Yen, IH and GA Kaplan. 1999. “ Neighborhood “The Contingent Meaning of Neighborhood social environment and risk of death: multilevel Stability for Residents; Psychologica l Well- evidence from the Alameda County Study.” being.” American Sociolog ical Review 65: American Journal of Epidemiol ogy 587–97. 151:1132–33. #1156—Jnl of Health and Social Behavior—Vol. 44 No. 4—44407-bro wning

MOVING BEYOND POVERTY 571

Christopher R.Browning is Assistant Professor of Sociology at the Ohio State University. His research interests include the effects of community structural disadvantage and social organization on risk behavior and health; the life-course consequences of childhood maltreament; and multilevel modeling techniques.

Kathleen A.Cagney is an Assistant Professor in the Department of Health Studies at the University of Chicago. Her research interests include access to and quality of long-term care arrangements, particularly race/ethnic and socioeconomic status differences; demography of aging, especially the role of family struc- ture in the timing of long-term care use; the effects of education and wealth on cognitive change at older ages; life course approaches to research in health and aging; neighborhood and health; social survey methodology and health status assessment.

Delivered by Ingenta to Welch Medical Library (cid 82007175), Johns Hopkins University (cid 90000553) IP: 127.0.0.1 Date: 2005..04..18..17..00..