See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254030611

Cultural Context, Sexual Behavior, and Romantic Relationships in

Disadvantaged Neighborhoods

Article in American Sociological Review · June 2007

DOI: 10.1177/000312240707200302

CITATIONS READS
179 187

1 author:

David James Harding

University of California, Berkeley
67 PUBLICATIONS 2,897 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

School Shootings View project

Incarceration and Prisoner Reentry View project

All content following this page was uploaded by David James Harding on 18 August 2015.

The user has requested enhancement of the downloaded file.

 National Poverty Center Working Paper Series

#06‐46

December, 2006

Cultural Context, Sexual Behavior, and Romantic

Relationships in Disadvantaged Neighborhoods

David J. Harding, University of Michigan

This paper is available online at the National Poverty Center Working Paper Series index at:
http://www.npc.umich.edu/publications/working_papers/

Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not
necessarily reflect the view of the National Poverty Center or any sponsoring agency.
 Cultural Context, Sexual Behavior, and Romantic Relationships

in Disadvantaged Neighborhoods*

David J. Harding

University of Michigan

December 15, 2006

*The author can be contacted at Population Studies Center, University of Michigan, 426 Thompson St., Ann Arbor,
MI 48106-1248, dharding@umich.edu, 734-763-2378 (phone), or 734-763-1428 (fax). Funding for this research was
provided by the National Science Foundation (SES-0326727), The William T. Grant Foundation, the American
Educational Research Association/Institute of Education Sciences, the MacArthur Foundation Network on
Inequality and Economic Performance, and by the Harvard Multidisciplinary Program on Inequality and Social
Policy, which is funded by an NSF Integrative Graduate Education and Research Traineeship grant. An NICHD
Post-Doctoral Fellowship at the Population Studies Center at the University of Michigan provided additional
support. Katherine Newman, Christopher Winship, Michele Lamont, Robert Sampson, Christopher Jencks, Jal
Mehta, Arland Thornton, and four ASR reviewers provided helpful comments on previous versions of this paper.
This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and
Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and
Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronald R.
Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from
Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-
2524 (addhealth@unc.edu).
 Cultural Context, Sexual Behavior, and Romantic Relationships

in Disadvantaged Neighborhoods*

Abstract

When culture is invoked to understand the consequences of growing up in disadvantaged

neighborhoods, the isolation of ghetto residents from mainstream institutions and mainstream
culture is often emphasized. This paper attempts to reorient current theorizing about the cultural
context of disadvantaged neighborhoods, particularly when it comes to adolescent decision-
making and behavior. It argues that rather than being characterized by the dominance of
“oppositional” or “ghetto-specific” cultures, disadvantaged neighborhoods are characterized by
cultural heterogeneity: a wide array of competing and conflicting cultural models. These ideas
are applied to sexual behavior and romantic relationships among adolescents using survey data
from Addhealth. Analyses show that disadvantaged neighborhoods exhibit greater heterogeneity
in cultural frames and scripts and that, in more heterogeneous neighborhoods, adolescents’
frames and scripts are poorly predictive of their actual behavior.

ii
 While sociologists have been concerned with urban residents and their neighborhoods

since the birth of the discipline (e.g. DuBois 1899, Park and Burgess 1925, Shaw 1929), the

publication of Wilson’s The Truly Disadvantaged (1987) and Massey and Denton’s American

Apartheid (1993) sparked a renewed interest among scholars of urban inequality in the role of

neighborhood context in the intergenerational transmission of poverty. These works directly

linked urban sociology to stratification, mobility, and race. In theoretical terms, neighborhoods

became an important context for the social processes driving stratification and racial inequality,

and neighborhood context was viewed as a causal force in the lives of youth and adults. These

works set off a sustained effort to understand the effects of neighborhood context on individual

outcomes, particularly for youth. Although there is mounting evidence that neighborhood effects

are real causal effects (see Harding 2003 for review and evidence on this issue), social scientists

have only begun to uncover the mechanisms by which such effects operate.

Culture has been largely ignored in the empirical effort to identify and understand the

social processes underlying neighborhood effects. Scholars studying race and poverty have shied

away from discussing culture since the early 1970’s, after critics of scholars such as Frazier

(1966), Moynihan (1965), and Lewis (1966, 1969) argued that cultural explanations of poverty

ignore structural barriers and blame the victim (e.g. Valentine 1968; see Wilson 1987 for a

review). In the interim, however, cultural sociology has moved away from the notion of culture

as an internally coherent set of values and toward a view of culture as fragmented and composed

of “disparate bits of information and … schematic structures that organize that information”

(DiMaggio 1997: 293). These ideas have only slowly found their way into the sociology of

disadvantaged neighborhoods. As I discuss further below, new concepts in cultural sociology

1
such as frames, scripts, and repertoires have the potential to illuminate the social processes at

work in neighborhood effects.

When culture is invoked to help us to understand the consequences of growing up in

disadvantaged communities, emphasis is often placed on the isolation of ghetto residents from

mainstream social networks and mainstream culture. Wilson (1987, 1996) argues that the out-

migration of the black middle class and the decline of manufacturing led to neighborhoods in

which life is no longer organized around work. Social interaction in isolated neighborhoods leads

to the development of cultural repertoires that are “oppositional” or “ghetto specific,”

adaptations to blocked opportunities in the labor market and society generally. Anderson (1990,

1991, 1999) invokes an alternative status system among adolescents from underclass

neighborhoods to understand high rates of teenage pregnancy and single parenthood. Massey and

Denton (1993) argue that racial segregation and the concentration of poverty lead to an

“oppositional culture” in inner cities in which conventional norms and values are upended in

response to blocked opportunities. “As intense racial isolation and acutely concentrated poverty

have continued, ghetto values, attitudes, and ideals have become progressively less connected to

those prevailing elsewhere in the United States. More and more, the culture of the ghetto has

become an entity unto itself, remote from the rest of American society and its institutions, and

drifting ever further afield” (Massey and Denton 1993: 172).

This paper is an attempt to reorient current thinking about culture in disadvantaged

neighborhoods and how it relates to adolescent decision-making and behavior. It re-introduces an

idea that was once a staple of theorizing about urban neighborhoods – disadvantaged

neighborhoods are characterized by cultural heterogeneity (Shaw and McKay 1969/1942, see

2
also Kornhauser 1978). 1 This paper applies these ideas, coupled with recent theoretical advances

in cultural sociology, to the analysis of adolescent sexual behavior and romantic relationships.

Using survey data from the National Longitudinal Study of Adolescent Health (Addhealth) on

frames regarding teenage pregnancy and scripts for romantic relationships, it shows that

disadvantaged neighborhoods exhibit greater heterogeneity in scripts and frames. This is the case

net of the greater ethnic and racial diversity and greater residential turnover that prevail in many

of these neighborhoods. This paper also investigates the consequences of cultural heterogeneity

for two adolescent behaviors, pre-marital sexual activity and the sequencing of events in

romantic relationships. Analyses show that in more heterogeneous neighborhoods it is more

difficult for adolescents to act in concert with the frames and scripts that they articulate.

I begin with a brief discussion of the role of cultural concepts in the dominant theories of

neighborhood effects. I then situate the concept of cultural heterogeneity in the urban sociology

and urban poverty literatures. Next, I describe the theoretical motivation for the cultural concepts

I use here and develop hypotheses about the causes of cultural heterogeneity in disadvantaged

neighborhoods and the consequences of this heterogeneity for adolescents. I then describe the

Addhealth data and the methods used to examine these hypotheses and show the results of these

analyses. The paper concludes by considering the implications for theories of neighborhood

effects.

Culture, Inequality, and Neighborhood Effects

Much recent work at the nexus of inequality and cultural sociology has focused on the

analysis of class. Symbolic boundaries have been used to understand how groups create and

1
Shaw and McKay did not use the term cultural heterogeneity but instead referred to “different systems of values”
and “different forms of organization.” As the term “different systems of values” suggests, they also employed a
more values oriented conception of culture than that used here.

3
maintain their separateness from and dominance over other groups (e.g. Lamont 1992, 2000;

Kefalas 2003; for a review see Lamont and Molnar 2002). Cultural capital theory has been used

to understand how the upper class employs cultural signals, credentials, and knowledge to

exclude others and to pass on advantages to children (Bourdieu and Passeron 1977, Bourdieu

1984, Lamont and Lareau 1988, DiMaggio 1982, Erickson 1996, Bryson 1996). Lareau (2000,

2003) has shown class differences in parenting practices and their role in educational outcomes.

Increasingly, poverty scholars have also employed cultural analysis, particularly in documenting

the cultural world of the urban poor, describing how they understand their options and make

decisions with regard to work, welfare, schooling, parenthood, and marriage (Newman 1999,

Anderson 1990, 1999, Waller 2002, Hays 2003, Young 2004, Carter 2005, Edin and Kefalas

2005). Yet culture has been less explicitly incorporated into the recent literature on the role of

neighborhood context in the intergenerational transmission of poverty.

Two mutually compatible theories have been offered to understand neighborhood effects

on individuals, and cultural concepts have been incorporated into them to different degrees.

Social organization theory argues that neighborhood disadvantage leads to difficulties

establishing and maintaining order. Lack of resources, racial and ethnic heterogeneity, and

population turnover result in fewer social ties and therefore diminished social control, the ability

of a community to regulate the behavior of its members (Park and Burgess 1925, Shaw 1929,

Shaw and McKay 1942). Communities with denser social networks are better able to articulate

and enforce common norms and values. In addition, local formal and informal institutions affect

the ability of neighbors to maintain social control by influencing norms and expectations and by

providing contexts within which social ties are created and strengthened. External institutions,

such as police, city government, and markets, affect the resources that are available for both

4
formal and informal methods of social control (Bursik and Grasmick 1993). Collective efficacy,

defined as the “social cohesion among neighbors combined with their willingness to intervene on

behalf of the common good,” mediates the relationship between concentrated structural

disadvantages (residential instability, ethnic or racial heterogeneity, and poverty) and crime rates

(Sampson et al.1997). Social organization models of neighborhood effects have thus far almost

exclusively been used to explain neighborhood differences in crime, violence, and delinquency

outcomes (exceptions include Browning, Leventhal, and Brooks-Gunn 2004, 2005, on

neighborhood collective efficacy and early sexual initiation).

Though social organization models are not usually thought of as cultural models, they do

incorporate cultural elements. In the classical formulation, social organization matters because

socially organized neighborhoods are better able to enforce common values. In addition,

collective efficacy can be thought of as a cultural concept, insofar as collective efficacy itself is

measured as residents’ commonly held expectations or beliefs about how others around them

will behave when faced with non-normative behavior. Finally, Small (2002, 2004) shows how

cultural frames regarding the neighborhood and its origins can impact residents’ willingness to

engage in the collective activity and institution building that lead to collective efficacy and social

organization.

While social organization theory focuses primarily on processes internal to the

neighborhood, social isolation theory emphasizes social and cultural disconnections between

neighborhood residents and the outside world. Social isolation theory argues that residents of

concentrated poverty neighborhoods are more likely to be isolated from middle class or

mainstream social groups, organizations, and institutions (Wilson 1987). The joblessness

endemic to high poverty areas means many residents are not connected to the mainstream labor

5
market, an important tie to the culture of middle class life (Wilson 1996). Lack of resources in

high poverty neighborhoods make sustaining neighborhood institutions more difficult, further

isolating neighborhood residents from associated mainstream institutions. The result is that social

interaction in isolated neighborhoods leads to the development of cultural repertoires that differ

from the mainstream. Youth are socialized into a cultural environment that promotes behaviors,

such as early sexual behavior and high school dropout, that are viewed as detrimental in the

outside world.

Most research on social isolation theory investigates the social connections of

neighborhood residents by examining whether neighborhood poverty is related to organizational

participation or to network ties to employed or college-educated individuals, net of individual

characteristics. Tigges, Browne and Green (1998) report that adults in higher poverty

neighborhoods in Atlanta have smaller “discussion networks” (people they talk to about

important matters) and discussion networks with fewer employed or college educated

individuals. Rankin and Quane (2000) find that neighborhood poverty predicts lower

organizational participation and fewer ties to employed individuals and to the college educated

and more ties to welfare recipients among adults surveyed in Chicago in 1991. Fernandez and

Harris (1992) report similar findings from Chicago adults surveyed in 1986 and 1987.

Meanwhile, the cultural predictions of social isolation theory have been left largely

uninvestigated. Social isolation theorists have relied heavily on the notion of “oppositional

culture” from ethnographic research on racial differences in educational performance by

Fordham and Ogbu (1986), extending the concept to apply in domains other than education (e.g.

Massey and Denton 1993). Fordham and Ogbu (see also Ogbu 2004) argue that poor black

students develop an oppositional culture in which behaviors that promote academic achievement,

6
such as speaking standard English, doing homework, and engaging in class discussion, become

defined as “acting white,” as a response to inferior schools, discrimination, and blocked

opportunities. 2 (See Willis 1977 and MacLeod 1995 on more class-based conceptions of

oppositional culture). However, survey research has rejected the claim that black students are

disproportionately sanctioned by their peers for academic effort (Cook and Ludwig 1998,

Ainsworth-Darnell and Downey 1998). These findings are reinforced in qualitative interviews

and survey research by Carter (2005), who shows that behaviors unconnected to school

achievement are at the heart of notions of “acting white” among poor black and Latino youth

(see also Jackson 2001). Moreover, cultural isolation and the development of a “ghetto-specific”

or “oppositional culture” in poor neighborhoods is further challenged by both survey-based and

ethnographic research on attitudes among the poor that finds very strong support for

conventional or traditional views about education, work, welfare, and marriage (Young 2004,

Newman 1999, 2003, 2006, Edin and Kefalas 2005, Solarzano 1992, Goldenberg et al. 2001;

Dohan 2003; Hayes 2003, Carter 2005, Waller 2002, Duneier 1992, Sayer, Wright, and Edin

2005). In short, there is little evidence that cultural isolation is an accurate description of the

cultural context of poor neighborhoods. This paper proposes a new conception of the cultural

context of disadvantaged neighborhoods, one which emphasizes the cultural heterogeneity of

such neighborhoods and the consequences of that heterogeneity for adolescent decision-making

and outcomes. 3

2
Fordham and Ogbu (1986) are often interpreted as arguing that academic success itself comes to be defined as
acting white within an oppositional culture, but in a recent paper, Ogbu (2004) clarifies that their argument was that
behaviors that lead to academic achievement are what are defined as acting white by poor black adolescents.
3
There are undoubtedly cultural dimensions on which middle class neighborhoods have comparable or greater
heterogeneity than poor neighborhoods, such as political views, fashion preferences, or religious beliefs. The focus
of this paper is on cultural dimensions related to individual outcomes typically studied by poverty and inequality
researchers, such as sexual behavior.

7
Cultural Homogeneity and Heterogeneity in Urban Neighborhoods

The concept of neighborhood cultural heterogeneity has roots in previous work in urban

sociology and is consistent with much ethnographic research on urban poverty. However,

cultural heterogeneity has been the subject of only limited explicit theorizing, perhaps because

much foundational work on urban neighborhoods rested largely on a view of neighborhoods as

culturally homogeneous. For example, Park and Burgess (1925) viewed urban neighborhoods as

immigrant receiving areas. Between neighborhood differences were the consequences of cultural

differences between immigrants’ home countries, and the culture that immigrant groups brought

with them was the basis of local neighborhood cultures (see also Gans 1962). These analyses,

like those of many of their contemporaries, view culture as a unified system.

Yet not all urban sociologists viewed neighborhoods as composed of homogenous

subcultures. Shaw and McKay (1969/1942) argued that socially disorganized slum

neighborhoods present youth with a wide array of “competing and conflicting moral values” both

conventional and unconventional, creating a breakdown of social control, which leads to higher

rates of delinquency in such neighborhoods. “In the areas of low rates of delinquents there is

more or less uniformity, consistency, and universality of conventional values and attitudes with

respect to child care, conformity to law, and related matters; whereas in the high-rate areas,

systems of competing and conflicting moral values have developed. Even though in the latter

situation conventional traditions and institutions are dominant, delinquency has developed as a

powerful competing way of life.” (Shaw and McKay 1969: 170).

Work in the tradition of urban ethnography has also complicated the stark divisions of

“ghetto culture” and “mainstream culture,” tending to see culture in disadvantaged

neighborhoods as derived from mainstream culture but modified or reinterpreted to serve local

8
needs and in response to blocked opportunities (e.g. Liebow 1967, Anderson 1978, Duneier

1992, Sullivan 1989, Bourgois 1995). Nevertheless, because these works focus on particular

groups within urban neighborhoods, there is little emphasis on cultural heterogeneity beyond

subjects’ attempts to distinguish themselves from culturally defined others lower in local status

hierarchies. Suttles (1968) emphasizes cultural differences within disadvantaged neighborhoods,

but he focuses on those between ethnic groups that use different communication devices and

have different cultural practices.

Hannerz (1969) was one of the first to recognize cultural heterogeneity not just across

groups residing together in a single neighborhood but also in the use of culture by individuals.

Though he draws on classic cultural concepts such as norms and symbolic meanings, he also

introduces the concept of cultural repertoire. For Hannerz, there are multiple forms of culture:

norms and values, meanings, and modes of action, and each individual has a repertoire of these.

Local cultures can add to or substitute for items in the mainstream cultural repertoire and thereby

provide adaptations and reactions to a given structural situation. Like Liebow (1967) and

Anderson (1978), Hannerz sees local culture as helping individuals to come to terms with

contradictions between the wider society’s culture and the individual’s position in the social

structure. Yet, Hannerz makes clear that “ghetto culture” is not a monolithic entity but rather a

heterogeneous mix of fluid ideal-type lifestyle groups (“mainstreamers,” “swingers,” “street

families,” and “street corner men”). In ghetto neighborhoods, members of these groups live in

close physical proximity, which often leads them to construct exaggerated social hierarchies and

distinctions (see also Newman 1992). Countering the divisive moral judgments between lifestyle

groups, however, are the family ties and spatial proximity that pull individuals with divergent

lifestyles into regular contact and confrontation.

9
 Recent research has also demonstrated that within disadvantaged neighborhoods, there

are multiple cultural models available. 4 For example, Newman (1999) shows that even in

neighborhoods with high levels of joblessness, the majority of people are pursuing activities

consistent with mainstream ideologies, such as working or going to school. Anderson (1990,

1999) documents the presence of both “street” and “decent” orientations among those living in

disadvantaged neighborhoods. Though they are in the numerical minority, those with a “street”

orientation dominate public space and public life in inner city neighborhoods.

The concept of cultural heterogeneity is also broadly consistent with classic research

traditions within the study of social stratification focusing on class inequality. Parkin (1971), in

his analysis of the normative order of the working class, describes this order as composed of a

number of competing meaning systems – the dominant, subordinate, and radical – each

“promoting a different interpretation of class inequality.” The result is a state of “normative

ambivalence,” in which the working class draws from a “reservoir of meaning” that is fed by

these three different “streams.” Rodman (1963) develops the concept of the “lower-class value

stretch” to explain how the lower-classes can have a somewhat distinctive value system that is

still derived from the dominant value system.

In sum, prior research suggests that the cultural context of disadvantaged communities

can be thought of as derived from mainstream culture but modified or reinterpreted to serve local

needs and in response to blocked opportunities. Most analysts recognize the presence of multiple

competing lifestyle groups (to use Hannerz’s terminology) or orientations (to use Anderson’s)

within urban neighborhoods. Culture within neighborhoods is not a single entity but rather a

4
I follow Quinn and Holland (1987) in my use of the term “cultural models,” which they define as “Presupposed,
taken-for-granted models of the world that are widely shared (although not necessarily to the exclusion of other,
alternative models) by the members of a society and that play an enormous role in their understanding of that world
and their behavior in it” (Quinn and Holland 1987: 4). I consider frames and scripts to be two types of cultural
models.

10
heterogeneous mix of lifestyles or orientations that individuals move between or draw upon as

necessary. Such a conception is not consistent with the cultural isolation of poor inner-city

neighborhoods from middle class culture and institutions.

However, analyses of the role of culture in explaining behavior in poor neighborhoods

still largely rest on social isolation models. These analyses tend to identify subcultures that

promote or justify particular behaviors as explanations for those behaviors. For example,

Anderson (1990, 1991, 1999) explains adolescent sexuality, gender relations, and teenage

pregnancy in disadvantaged neighborhoods with the development of a subculture in which early

sexual activity and early parenthood are normative. The coexistence of these two incompatible

models (one that describes poor neighborhoods as containing a mix of non-discrete cultural

groups and another that relies on cultural subgroup explanations of behavior), creates an

analytical conundrum: If individuals draw from multiple cultural lifestyle models, how can

subcultures hold such sway over behavior, action, or decision-making? I propose that recent

efforts to incorporate new cultural concepts such as frames, scripts, or repertoires into theorizing

about culture in urban neighborhoods are an important start toward better understanding the

relationship between culture and behavior among adolescents in disadvantaged neighborhoods.

Conceptualizing Cultural Heterogeneity

In the last two decades cultural sociology has moved away from a view of culture as an

internally coherent set of values and toward a view of culture as fragmented and composed of

“disparate bits of information and … schematic structures that organize that information”

(DiMaggio 1997: 293). In any social context, from the perspective of any individual, there are

multiple cultural models available from which to choose (Swidler 1986, Quinn and Holland

11
1987, Holloway et al. 1997, Fuller et al. 1996). These models may be overlapping or

contradictory, and reflect ideas about how the world works, what appropriate goals are, and how

to go about accomplishing things. 5 To further unpack these issues, I rely on three

conceptualizations, culture as repertoire or toolkit (Hannerz 1969, Swidler 1986, 2001), culture

as frame (Goffman 1974, Benford and Snow 2000, Small 2002, 2004), and culture as script.

Swidler (1986, 2001) draws upon the concept of cultural repertoire to develop a general

conception of culture that allows culture to play a causal role in influencing action. 6 Swidler sees

culture as a “tool kit” of symbols, stories, and worldviews which people use to solve different

problems. Under this model, culture is not a unified system but a repertoire from which to draw.

Culture provides the components used to construct “strategies of action,” or “persistent ways of

ordering action through time” and can thereby have a causal role. The elements that make up

one’s toolkit come not just from direct experience or social interaction but also from the wider

culture through institutions such as the media, schooling, and religion. The ability of culture to

predict behavior in the tool kit model comes from variation in repertoires across cultural groups

or across individuals. 7

In the analysis below, I measure two types of cultural objects that may be present in an

individual or group’s cultural repertoire: frames and scripts. Frames are ways of understanding

the world and how it works. They encode expectations about consequences of behavior, and how

various parts of the social world relate or do not relate to one another. A frame impacts how we

5
A conception of culture as heterogeneous and contradictory appears elsewhere as well. For example, Sewell notes
that, “social actors are capable of applying a wide range of different and even incompatible schemes” (Sewell
1992:17). He also describes cultures as contradictory, loosely integrated, contested, subject to constant change, and
weakly bounded (Sewell 1999).
6
Tilly (1978) is also credited with the development of the concept of repertoire in his work on “repertoires of
collective action.”
7
Swidler’s toolkit metaphor has been criticized for failing to specify how individuals choose which elements of their
toolkit to employ in different situations (Lamont 1992, Lamont and Thevenot 2000). Lamont (1992) suggests that
both proximate and remote structural conditions influence such selections. For an example of comparison of
repertoires across cultural groups, see Lamont and Thevenot (2000).

12
interpret events and therefore how we react to them. Benford and Snow (2000), writing from a

social movements perspective, emphasize that frames are collectively constructed, often

unintentionally but sometimes intentionally. They identify problems and assign blame, provide

solutions or strategies, and provide a rationale for engaging in action. Like repertoires, frames

allow for cultural heterogeneity. Individuals can have multiple contradictory or competing

frames which they deploy in different situations, and frames may have various levels of

specificity. Small (2002, 2004) shows that individuals in the same neighborhood can employ

different frames. He develops a variant of collective action frames that he calls “neighborhood

frames” to explain the decline across cohorts in participation in community activities in a Puerto

Rican housing project. Such frames are the product of historical conditions experienced by

different cohorts. Small also shows that frames are not fixed, as young people’s neighborhood

frames change through interaction with the neighborhood activists of the earlier generation.

Scripts provide cultural templates for the sequencing of behaviors or actions over time.

They are akin to Swidler’s “strategies of action” in that they show how to solve problems or

achieve goals. There need not be consistency across various scripts or frames as individuals are

often able to live with many contradictions and inconsistencies. Therefore, one should not think

of frames and scripts as necessarily hierarchically nested, and individuals or groups may possess

or employ multiple contradictory frames and scripts. What people of a common culture share,

then, is not a coherent, monolithic culture but a set of available frames and scripts, objective

structural conditions, and knowledge of what others do and think. Sewell (1999) characterizes

this level of coherence as “thin coherence.”

Though I have conceptualized cultural heterogeneity using what might be described as

macro or meso-level cultural concepts, it is also consistent with micro or interactionist

13
perspectives on culture. Implicit in Goffman’s (1974) analysis of frames is the notion that

multiple frames can apply in any interaction, and that participants must work to maintain the

dominance of a particular frame and often come into conflict over which frame should govern a

particular situation (see also Goffman [1959]). His discussion of “out of frame activity”

acknowledges that multiple frames can be present simultaneously, even if they are not mutually

compatible. The cultural heterogeneity approach also shares with interactionist approaches the

criticism that subcultural analyses tend to assume homogeneity and stasis and tend to define

culture in terms of values (Fine and Kleinman 1979). Furthermore, interactionist perspectives

emphasize the role of local or group cultures in mediating the relationship between the wider

social environment and individual action (e.g. Fine 1979). As will be discussed below, the

consequences of neighborhood cultural heterogeneity occur in part through micro-level

interactions with neighbors with varying cultural orientations or lifestyles and through local

interpretations of cultural information diffused through the media.

Neighborhood Disadvantage and Cultural Heterogeneity

Why are adolescents in disadvantaged neighborhoods presented with a more

heterogeneous array of cultural models from which to fashion their beliefs and actions than those

in more advantaged neighborhoods? Among disadvantaged neighborhoods, all but the most

extremely poor neighborhoods contain a mix of people who are working and people on public

assistance or involved in crime, individuals who are high school dropouts and individuals who

have attended college, and families that have middle class incomes and families that are

struggling below the poverty line. As Patillo-McCoy (1999) notes, though many middle class

14
blacks left inner city neighborhoods in the 1970’s, many also remained there. This mix means

that adolescents will come into contact with people with a wide array of lifestyles. 8

In addition, cultural models do not just come from immediate ecological factors or

interpersonal interactions. Institutions that are decoupled from everyday interactions such as the

media, religion, and politics also contribute to cultural repertoires and are another mechanism by

which youth in disadvantaged neighborhoods are exposed to mainstream or middle class culture.

For example, youth draw role models from television and radio (Carter 2005: Ch. 5); the

American dream provides a cultural template for many of the poor (Hochschild 1995); feminism

shapes how young mothers think about economic independence and marriage (Edin and Kefalas

2005); and religion provides a repertoire for constructing strategies of action in black

communities (Patillo-McCoy 1998).

Residents of disadvantaged neighborhoods are also often less able to control public

behavior or regulate those who deviate from mainstream lifestyle choices. There is weaker

consensus on which behaviors are appropriate and inappropriate. A lack of strong social ties in

such neighborhoods means social control is diminished as local institutions are weak and

collective behavior is more difficult (Shaw and McKay 1969). Sampson et al. (1997), for

example, show that low collective efficacy in poor Chicago neighborhoods accounts for much of

the increased violence that these neighborhoods experience. Residents of poor neighborhoods

lack confidence that their neighbors will intervene to stop public disorder and so hesitate to do so

themselves. Such breakdowns of public order can be self-perpetuating, as residents lose touch

with neighbors as they retreat from the streets, leading to further weakening of neighborhood

8
There are of course many factors which determine the capacity of culture to influence behavior. Schudson (1989)
outlines five “dimensions of cultural power:” retrievability, rhetorical force, resonance, institutional retention, and
resolution. Here I emphasize retrievability, the availability or accessibility of a cultural model or cultural object.
Schudson suggest that retrievability is highest when a cultural model or object is physically present, institutionalized
in common practice or public memory, or more salient because it is more recent or more dramatic.

15
social ties (Venkatesh 2000). Similarly, Wilson (1996) argues that the lack of social organization

in these neighborhoods makes “ghetto-related behaviors” more acceptable, while the lack of

opportunities makes them necessary for survival. As the number of individuals pursuing

“ghetto-related” behaviors such as early parenthood, reliance on public assistance, or street

hustling increases, the stigma attached to these behaviors declines.

Of course the mere presence of a mix of mainstream and alternative lifestyles in

disadvantaged communities does not necessarily mean that adolescents will interact with

individuals who engage in these various lifestyles. In disadvantaged neighborhoods, lifestyles

outside the mainstream may have more salience than sheer numbers would lead us to expect. In

contrast to those with mainstream lifestyles who leave the neighborhood for work or school,

those with “street orientations” have a daily presence in the neighborhood. Anderson (1999)

documents the power of drug dealers and others who use violence to control public space and

flaunt material goods such as cars, jewelry, and clothing. While ethnographers such as Suttles

(1968) and Hannerz (1969) have described the strong divisions along ethnic and lifestyle lines

that occur when groups in physical proximity define themselves as fundamentally different or

better than one another, today’s young people may be more forgiving of racial and ethnic

differences than those of the 1960’s (see, for example, Carter 2005). In addition, Hannerz’s

neighborhood hierarchies are created and maintained largely by adults, not adolescents. Indeed,

he describes how “mainstream” parents go to great pains to keep their children away from “street

families” (see also Furstenberg et al. 1999).

These factors related to neighborhood disadvantage suggest the first hypothesis that this

paper will examine:

16
 Hypothesis 1: Adolescents in more disadvantaged neighborhoods will exhibit greater

heterogeneity of cultural frames and scripts.

The Consequences of Culturally Heterogeneous Neighborhoods

Adolescents may be particularly influenced by cultural heterogeneity because their

developmental life stage is associated with a strong focus on social identities and peer influences

as a well as with greater experimentation and risk taking. What are the consequences for

adolescents who face a wider array of cultural models in their social environment? If an

individual’s repertoire is constructed from what he observes among the people with whom he

interacts and from the broader cultural ideas to which he is exposed through media and

institutions, then the average adolescent in a disadvantaged neighborhood will have a wider

range of scripts and frames in his repertoire. From an interactionist perspective (e.g. Fine 1979),

significant cultural fluidity is created when cultural models diffuse across small groups through

local (neighborhood) and extended interaction networks.

Consider the multiple frames around teenage pregnancy that exist in disadvantaged

neighborhoods and inform adolescents about the consequences of becoming a father or mother at

a young age. One frame defines early parenthood as a hindrance to schooling and future success.

Another frame highlights the status that comes from parenthood and the attention and admiration

that a new baby brings (Anderson 1990, 1999). A third frame implies that having a child too late

in life will mean one will be too old to have the energy to enjoy the child and to be a good

parent. A fourth highlights the power of a child to boost a young man’s masculinity or a young

woman’s femininity or the power to instantly grant adult status to a new parent (Anderson 1990,

1999).

17
 Consider also the multiple scripts for romantic relationships that are available to an

adolescent in a disadvantaged neighborhood. One such script may follow a conventional path

starting with casual dating, the exchange of gifts, exclusive dating, statements of love, sexual

activity, marriage, living together, and pregnancy. Another script may involve little dating and a

fast transition to sexual activity and cohabitation, followed by pregnancy, and far in the future,

possibly marriage (Edin and Kefalas 2005). A third may involve early statements of love to

multiple partners, non-exclusive dating, and sexual activity with multiple partners and the

potential for unplanned or at least unavoided pregnancy (Anderson 1990, 1991, 1999).

When it comes, then, to choosing a course of action from among available relationship

scripts or to considering the various frames that define the risks and benefits of a particular

behavior such as teenage sexual activity, adolescents in disadvantaged neighborhoods have much

more to consider than those in more affluent areas. Not only do disadvantaged neighborhoods

provide a wider array of models, but – unlike more middle class communities – there is also

social support for both mainstream and alternative models. One can observe neighbors engaging

in behaviors that conform to various cultural models. In addition, when there are multiple

models, the advantages and disadvantages of various options are more poorly defined. The social

environment provides a much weaker signal about what option is best because there is social

support from others for many different options. We can imagine two consequences of this

problem.

First, because there is less consensus or agreement where there is greater heterogeneity, it

will be harder for the adolescent to choose between competing options. The result may be

weaker commitment to the chosen option and a lower likelihood of follow-through. With a

weaker commitment, when a particular option does not seem to be working out, it is easier to

18
shift course because another option is available – with local approval. Obstacles or setbacks may

be more likely to push the adolescent off track. For example, when a romantic partner is not

serious about contraception, one can downplay the “pregnancy as roadblock” frame and adopt

the “parenthood as status” frame to understand one’s options and behaviors. There are always

others one can look to who appear to be surviving or even thriving while engaging in behavior

consistent with various cultural models.

The second consequence arises at the aggregate level. In a social environment in which

many options are present and visible, there will be fewer who have enacted a particular option.

Because fewer others have followed a particular script, there will be fewer examples of how to

do so. In other words, the details of the available scripts will be less clearly defined, which may

make it more difficult to successfully enact the script. On the one hand, this can be thought of as

an information problem. For example, when fewer neighbors have successful long-term romantic

relationships, how to go about developing such a relationship will be less clearly defined. On the

other hand, it can also be thought of as a social support problem, in that in a culturally

heterogeneous environment there will be considerably less than unanimous social and cultural

support for particular frames or scripts. For instance, one’s frame about the advantages and

disadvantages of early parenthood may face frequent challenge from others with different

frames.

These arguments suggest that in the context of neighborhood cultural heterogeneity,

adolescents will be less likely to act in ways consistent with frames and scripts that they

articulate. 9 This leads to the second hypothesis that the analysis below will examine:

9
These arguments are also consistent with the social psychological literature on the link between attitudes and
behavior (though note that frames and scripts are conceptually different from attitudes, which social psychologists
define as summary evaluations of the degree of overall favorability toward a psychological object [Ajzen 2001]).
According to the Fishbein/Ajzen attitude behavior model, discrepancies between attitudes and behavior can be

19
 Hypothesis 2: In culturally heterogeneous neighborhoods, there will be a weaker

relationship between the script or frame that an adolescent has adopted and his or her

corresponding future behavior.

Methods

Data

I use data from the National Longitudinal Survey of Adolescent Health (Addhealth;

Harris et al. 2003). The Addhealth survey initially sampled about 150 middle schools, high

schools, and junior high schools. The first wave of data was gathered in 1994-1995, the second

wave in 1996, and the third wave in 2001-2002. Students were in grades 7 to 12 in wave one.

The first wave of data includes a short questionnaire completed by school administrators about

school characteristics and policies, an in-school questionnaire completed by almost every eligible

student in the sample schools, and longer in-home student and parent interviews with a

subsample of about 20,000 students. Wave two followed the in-home students and includes

another in-home interview with the student (but not the parent) and another school administrator

questionnaire. Structural neighborhood characteristics from the 1990 census are available for in-

home respondents. The strengths of the Addhealth data for this analysis are the nesting of

respondents within neighborhoods, the availability of neighborhood structural characteristics,

and the availability of survey measures of frames and scripts.

produced by (among other things) moderating third variables, for example, group norms (Fishbein and Ajzen 1975;
for review and critique, see Ajzen 2001, Liska 1984). In another example, Barber (2001) finds that the relationship
between attitudes toward children and marital childbearing behavior is affected by attitudes toward competing
behaviors such as careers and material consumption. Attitude-behavior discrepancy can also be a product of attitude
specificity. For instance, concrete attitudes, those derived from personal and family experiences, predict black
adolescents educational behavior, whereas abstract attitudes do not (Mickelson 1990). Here, neighborhood cultural
context serves as a moderator of the relationship between cultural models, such as frames or scripts, and behavior.

20
Models

To examine Hypothesis 1, that disadvantaged neighborhoods exhibit greater cultural

heterogeneity, I constructed neighborhood level measures of cultural heterogeneity for two

cultural concepts, pregnancy frames and romantic relationship scripts (described below). I then

regress each measure of neighborhood heterogeneity on neighborhood disadvantage, controlling

for other neighborhood characteristics that might also produce cultural heterogeneity, such as

racial and ethnic diversity or residential instability. More disadvantaged neighborhoods are

expected to have greater heterogeneity.

To examine Hypothesis 2, that adolescents in more heterogeneous neighborhoods are less

likely to act in accordance with their articulated frames and scripts, I estimate a series of multi-

level regression models, one set for pregnancy frames and one set for relationship scripts. The

pregnancy frame model examines the relationship between pregnancy frames (F) and engaging

in sexual activity (Y) using an interaction between individual pregnancy frame and neighborhood

heterogeneity in pregnancy frame. If we index individuals with i, neighborhoods with j, and

schools with k, we can write a three-level model (Raudenbush and Bryk 2002). 10 The individual

level model is:

⎛ Pr(Yijk = 1) ⎞
log⎜ ⎟ = π 0 jk + π 1 jk Fijk + π 2 jk X ijk + eijk
⎜ 1 − Pr(Y = 1) ⎟
⎝ ijk ⎠

Because Y is binary, I use a logit link. F is a measure of the individual’s pregnancy frame

(described below), and X is a set of control variables measuring individual and family

characteristics (and π2 is a vector of coefficients). There are two neighborhood level equations:

10
Though schools are not of analytical interest here, school is included as a level in the model because of the
structure of the data and to allow school characteristics to be used as control variables.

21
 ~
π 0 jk = β 00 k + β 01k F jk + β 02 k W jk + r jk
~
π 1 jk = β 10 k + β 11k F jk + β 12 k V jk

The first equation models the intercept from the individual level model as a function of
~
neighborhood heterogeneity in the pregnancy frame (F ) and a set of neighborhood control

variables (W), including the neighborhood mean pregnancy frame. The second equation can be

thought of as adding cross-level interaction terms. It models the coefficient capturing the

relationship between the individual frame (F) and sexual activity (Y) from the individual level
~
model as a function of neighborhood heterogeneity in the pregnancy frame (F ) and a set of

neighborhood control variables (V), again including the neighborhood mean pregnancy frame. It
~
is the coefficient on F , β11k , in the second neighborhood level equation that tests Hypothesis 2

in this model. Finally, there is a school level equation that serves to control for a set of school

characteristics, Z:

β 00 k = γ 000 + γ 001 Z k + u k

The analysis proceeds by estimating a set of increasingly more complicated models that build up

to the model just described. Earlier models set various control variable coefficients to zero

(effectively leaving out the control variables).

The romantic relationship scripts model is somewhat simpler. In this model I measure

directly the difference between one’s ideal romantic relationship and one’s subsequent actual

romantic relationship (described in more detail below), denoted Y. This is the dependent variable.

The primary predictor of interest is neighborhood heterogeneity in ideal romantic relationship

~
scripts (also described in more detail below), denoted R . Again we can write a three level

model. The individual level equation is:

22
 Yijk = π 0 jk + π 1 jk X ijk + eijk

The difference between one’s ideal and actual romantic relationship script is modeled with an

intercept and a series of control variables for individual and family characteristics (X). The

neighborhood level equation is:

~
π 0 jk = β 00 k + β 01k R jk + β 02 k W jk + r jk

This equation models the individual level intercept as a function of neighborhood heterogeneity
~
in ideal romantic relationship scripts ( R ) and neighborhood control variables (W). The
~
coefficient on R , β01k, is the test of Hypothesis 2 for relationship scripts. It indicates the degree

to which neighborhood heterogeneity in relationship scripts predicts the difference between an

individual’s ideal relationship script and his or her subsequently realized relationship. The school

level equation serves to control for a set of school characteristics, Z:

β 00 k = γ 000 + γ 001 Z k + u k

Again, the analysis proceeds by building up to a model with the full set of neighborhood and

school control variables.

Variables

Structural Neighborhood Disadvantage: Neighborhoods are measured as census tracts. As

is the convention in neighborhood effects research (e.g. Sampson et al. 1997), neighborhood

disadvantage is measured by a scale constructed from series of neighborhood structural

characteristics. Here the neighborhood disadvantage scale is the mean of the following

standardized items: the census tract’s family poverty rate, percent single mother households,

percent youth, male unemployment rate, percent black, percent of those over 25 who are college

graduates, percent of workers in managerial or professional occupations, and percent affluent

23
families (those with incomes above $75,000 per year), with the last three reversed in polarity.

These data come from the 1990 census. An individual’s census tract is that of his or her

residence at the wave one in-home interview, which was conducted in spring or summer of 1995.

The average inter-item correlation for this scale is 0.52 and Cronbach’s alpha is 0.90.

The Structural Neighborhood Disadvantage Scale (hereafter, Neighborhood

Disadvantage) measures the economic and social characteristics of the families that make up the

neighborhood and which are thought to lead to negative outcomes for youth. 11 Five of these

variables (poverty, single mother households, percent youth, male unemployment, and percent

black) indicate the presence of disadvantaged families. 12 The remaining three (college graduates,

managerial and professional workers, affluent families) indicate the absence of middle class

families since their polarity is reversed. While some researchers (e.g. Brooks-Gunn et al 1993)

have argued that the absence of middle class families is more important than the presence of

disadvantaged families, there are high inter-item correlations across all eight variables in these

data. This suggests that these two sets of measures capture the same underlying neighborhood

SES concept but simply focus on the presence of families at opposite ends of the SES

distribution as indicators of a neighborhood’s position in that distribution. 13

Pre-marital Sex: Pre-marital sex between wave one (when almost all respondents are age

13 to 18) and wave two (about a year later on average) measures whether the adolescent reports

11
Neighborhood disadvantage is not meant to directly proxy for neighborhood cultural context as is sometimes done
when direct measures of cultural context are not available. One of the goal’s of this paper is to investigate the
relationship between structural, and therefore cultural characteristics of neighborhoods and cultural characteristics
are measured directly, as described below.
12
Percent youth roughly captures the number of adults per child to supervise or monitor young people, both within
families and in the neighborhood in general.
13
A correlation matrix of these variables is available from the author upon request.

24
having had sexual intercourse outside of marriage between the first and second waves of the

survey. 14

Romantic Relationship Scripts: The Addhealth in-home data contain a series of questions

about adolescents’ ideal and actual romantic relationships. Ideal relationships are measured at

wave one by asking the adolescent to choose the events that would be part of his or her ideal

romantic relationship and then to arrange the events in his or her preferred chronological order.

For the ideal relationship scripts, respondents choose from and then order the following list of 17

events:

• We would go out together in a group

• I would meet my partner’s parents
• I would tell other people that we were a couple
• I would see less of my other friends so I could spend more time with my partner
• We would go out together alone
• We would hold hands
• I would give my partner a present
• My partner would give me a present
• I would tell my partner that I loved him or her
• My partner would tell me that he or she loved me
• We would think of ourselves as a couple
• We would talk about contraception or sexually transmitted diseases
• We would kiss
• We would touch each other under our clothing
• We would have sex
• My partner or I would get pregnant
• We would get married

In wave two, each respondent is asked to describe up to three past romantic relationships using

the same procedure. However, in wave two the menu of possible events is slightly different. The

14
Researchers studying adolescent sexual behavior from a more demographic perspective often focus on the
transition to first sex rather than sexual activity during a given time period. In this case, because of the emphasis on
the effect of neighborhood context, conditioning on prior sexual behavior would be conditioning on an endogenous
variable, one that is also affected by neighborhood context. In addition, in this case limiting the sample to those who
never had sex prior to wave one would result in a large reduction in statistical power, which is an issue in this
analysis because of the importance of interaction terms. Models estimated on this more limited sample show
substantively similar results but with larger standard errors (models not shown).

25
two events involving telling each other that you love one another are collapsed into one event in

wave two. The two events involving giving and receiving gifts are also collapsed into one event

in wave two. Getting married is not included in the menu of events in wave 2, and an additional

event, “You touched each others genitals (private parts),” is added in wave two. I consider the

ideal relationship from wave one to be a measure of the relationship script for the respondent’s

ideal romantic relationship. I then use the first relationship listed at wave two that started after

the wave one interview as the actual relationship script.

The analysis requires constructing two measures, one that captures heterogeneity in ideal

relationship scripts within a neighborhood and another that captures the degree to which an

individual respondent’s actual relationship script differs from his or her ideal relationship script.

Hypothesis 2 predicts that adolescents in neighborhoods with more heterogeneity in ideal

relationship scripts will have actual relationships that are more different from their ideal

relationship scripts.

The string of events in the relationship scripts can be thought of as sequence data, and

sequence analysis methods can be used to measure the resemblance or difference between two

scripts. 15 Optimal matching provides a method for comparing two sequences based on the

insertions, deletions, and substitutions required to transform one sequence into the other (Abbott

1990, 1995, Abbott and Hrycak 1990, Sankoff and Kruskal 1983; see also Stovel, Savage, and

Bearman 1996). Each insertion, deletion, or substitution carries a cost, and the optimal matching

algorithm finds the minimum total cost to transform one sequence into the other. This cost is a

15
I thank Hannah Brueckner and Peter Bearman for suggesting the use of sequence analysis methods and for advice
on their implementation.

26
measure of the difference between two sequences. This analysis uses the implementation of

optimal matching in the software Transition Data Analysis (Rohwer and Poetter 2005). 16

To measure neighborhood heterogeneity in ideal relationship scripts, I take the mean of

the differences between ideal romantic relationship scripts for all possible pairs of individuals in

each neighborhood (defined as a census tract). 17 The reliability of this measure depends heavily

on the number of Addhealth respondents in the neighborhood, so all models are weighted by this

reliability. 18 Further information on the measurement of differences in romantic relationship

scripts using sequence analysis is provided in Appendix C.

Pregnancy Frames: The degree to which a respondent agrees or disagrees with the

statement, “It wouldn’t be all that bad if you got pregnant [or got someone pregnant] at this time

in your life,” (see Table 1 below) measures his or her frame regarding teenage pregnancy. Above

I conceptualized frames as ways of thinking about the world and about the consequences of

one’s actions. While ideally one would use a measure that distinguished between multiple frames

regarding pregnancy (such as early parenthood as status, early parenthood as hindrance, or early

parenthood as leading to a closer relationship between parent and child), no such measures are

available in these or any other survey data. However, this survey question is a reasonable

measure of the respondent’s frame regarding teenage pregnancy because it captures both the

16
Previous sociological applications using sequence analysis have investigated sequences of jobs within careers
(e.g. Abbott and Hrycak 1990, Stoval et al. 1996). These applications involve sequences that explicitly take into
account time. In the Addhealth relationship scripts there is no information about the elapsed time between events in
the sequence, but there is nothing inherent to the optimal matching algorithm that requires information on elapsed
time. The measures constructed here thus focus particularly on the ordering of events in a romantic relationship.
This analysis also differs in the purpose of the sequence analysis. Previous applications typically calculate the
difference between all possible pairs of sequences and then use cluster analysis to determine a set of typical
sequences. Here I am using optimal matching only to measure particular characteristics of individuals and
neighborhoods.
17
Thus neighborhood heterogeneity in relationship scripts is undefined for neighborhoods with only one Addhealth
respondent (or only one Addhealth respondent with complete ideal relationship script data).
18
If nj is the number of respondent pairs in neighborhood j, σ2u is the within neighborhood variance, and σ2e is the
between neighborhood variance, the reliability of the neighborhood mean is σ2u/(σ2u + σ2e/ nj). Supplemental
analyses (not shown) indicate that models in Table 4 are not sensitive to the exclusion of tracts with low
measurement reliability.

27
valence and strength of an adolescent’s thinking about the consequences of teenage pregnancy.

By including the phrase, “at this time in your life,” it cues respondents to think of consequences,

and in particular of the consequences for themselves rather than teenagers in general. Because of

this practical orientation, this measure can be distinguished from more general attitudinal

measures that tap values regarding the moral status of teenage sexual behavior or early

parenthood. 19 20

To measure an individual’s adherence to this “pregnancy would have negative

consequences for me” frame, I assign responses the numbers one through five such that higher

values indicate a stronger feeling that pregnancy would have negative consequences. Calculating

neighborhood heterogeneity in pregnancy frames is complicated by the ordinal nature of this

measure, which makes the usual measure of variation for interval measures, the variance,

inappropriate. I use a measure of ordinal variation developed by Blair and Lacy (2000) that

measures concentration:
k −1

∑ (F i − .5) 2
l2 = i =1

(k − 1) / 4

k is the number of categories (in this case k = 5), and Fi is the cumulative proportion for category

i (i.e., Fi = ∑ j =1 p j where pj is the sample proportion for the jth of k categories). The numerator
i

measures the difference between the observed distribution and a distribution with maximum

19
For example, contrast this survey question with survey questions in the General Social Survey measuring attitudes
toward pre-marital sex or sexual activity by teenagers: “There's been a lot of discussion about the way morals and
attitudes about sex are changing in this country. If a man and woman have sex relations before marriage, do you
think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?” and, “What if they are
in their early teens, say 14 to 16 years old? In that case, do you think sex relations before marriage are always
wrong, almost always wrong, wrong only sometimes, or not wrong at all?”
20
This interpretation is also based on qualitative interviews with 60 adolescent boys in three neighborhoods in
Boston in which this survey question was administered to each boy and then both his understanding of the question
and his response were discussed (these data are discussed in more detail in [author cite removed]). When discussing
how they thought about answering this question, the boys focused on how their lives would be different were they to
become a parent in the near future and how that would affect their future plans and life chances.

28
dispersion, which occurs when responses are evenly divided between the two opposite extremes

(the minimum dispersion occurs when all responses are in a single category; see Blair and Lacy

2000). The denominator normalizes by dividing by the maximum possible value of the

numerator, so that l2 varies from 0 to 1. Because some neighborhoods have small numbers of

adolescents, Blair and Lacy’s small sample bias-adjusted l2 is required:

1− l2
l u2 = l 2 −
N −1

Finally, I take 1- l2u as my measure of neighborhood heterogeneity because l2u is a measure of

concentration. 21 Finally, the regression models also require controls for central tendency of the

distribution of frames in neighborhoods, for which I use the mean. 22

One complication in constructing the neighborhood heterogeneity measure by

aggregation of individual respondents to the neighborhood level in the Addhealth data is the

small number of respondents in some neighborhoods. For about half of the tracts in which at

least one Wave I Addhealth respondent lives there are no other Addhealth respondents, and

therefore it is impossible to measure heterogeneity for such tracts. These tracts are dropped from

the analysis. However, since most Addhealth respondents live in tracts with many other

respondents, this results in the loss of only about six percent of respondents from the multi-level

models predicting individual outcomes. For another 25 percent of tracts, there are less than five

respondents per tract, leading to low reliability of neighborhood level measures created by

21
Ideally, one would construct gender specific measures of neighborhood cultural heterogeneity of both pregnancy
frames and relationship scripts. Unfortunately, doing so with these data would cut in half the number of individuals
per neighborhood who contribute to the neighborhood-level measures, significantly reducing the reliability of such
measures.
22
Strictly speaking, the mean is not an appropriate measure of central tendency for an ordinal variable because it
assumes that the distances between categories carry information, whereas with an ordinal variable, only information
about the order of the categories is present. However, other potential measures of central tendency such as the mode
or median result in little variation across neighborhoods, and so would not serve as effective controls for central
tendency when measuring the impact of neighborhood heterogeneity.

29
aggregation. I weight models by the reliability of the mean to account for the low reliability of

neighborhood measures when they are constructed from small numbers of respondents.23

Neighborhood Controls: In models that examine the relationship between neighborhood

disadvantage and neighborhood cultural heterogeneity, it is necessary to control for other

potential sources of neighborhood heterogeneity. Social organization theory points to three other

potential sources of heterogeneity: immigration, residential instability, and ethnic and racial

diversity. To control for immigration, I use the percent of neighborhood residents who are

foreign born. To control for residential instability, I control for percent of housing units that are

owner occupied and percent of housing units that have housed the same household for the last

five years. 24 To control for ethnic diversity, I control for percent Hispanic and percent Hispanic

squared. Finally, to control for racial diversity, I use Simpson’s Interaction Index (White 1986,

cited in Reardon and Firebaugh 2002). This index is constructed from the percent of

neighborhood residents in four racial groups: White (W), Black (B), Asian (A), and Other (O):

⎛ W B A O ⎞
Racial Diversity Index = 100⎜ + + + ⎟
⎝1−W 1− B 1− A 1− O ⎠

This index varies from zero (entire neighborhood population from one group) to 75

(neighborhood population evenly divided between the four groups). 25

Individual/Family Controls: To minimize selection bias due to differences between

individuals across neighborhoods, individual and family control variables are required. Measured

at wave one, these controls include race and ethnicity indicators, age, gender, adolescent

23
Supplemental analyses (not shown) indicate that results in Table 3 are not sensitive to the exclusion of tracts with
low measurement reliability.
24
These two variables are obviously somewhat collinear, but their role in the model is to control for potential
confounding due to residential instability. I am not interested in the magnitude or statistical significance of their
coefficients.
25
I also experimented with Thiel’s Entropy Index (Thiel 1972, cited in Reardon and Firebaugh 2002). This index
sums p(ln(1/p)) for each of the four groups. The Simpson index was a better predictor of neighborhood
heterogeneity.

30
immigrant status, language spoken at home, log family income, single parent household, step-

parent or other household, mother’s age at birth, low birth weight, and for the primary parent

(mother or female caregiver if available, father or male caregiver if not) immigration status,

education, professional/managerial occupation, disability, and welfare receipt. These variables

are described in more detail in Appendix A.

School Controls: Since previous research has found that school characteristics are

important predictors of sexual and romantic behavior (e.g. Teitler and Weiss 2000), I include

school controls in the multi-level models. School controls include indicators for private school,

Catholic school, and rural/suburban/urban, and measures of school size, percent of students in a

college prep program, and cumulative dropout rate. These variables are also described in more

detail in Appendix A.

Several individual, family, and school control variables have missing values. Rather than

dropping cases with missing values on control variables, I impute missing values using multiple

imputation, currently believed to be among the best methods for dealing with missing data (see

Acock 2005 for a non-technical discussion and references therein for more technical material,

such as Little and Rubin 2002, Allison 2002). Multiple imputation involves creating multiple full

datasets via MICE (multiple imputation by chained equations), estimating a model using each

full dataset, and then combining results across datasets in a way that takes into account the

variance in imputed values across datasets.26 Here I use ten imputed datasets. Finally, continuous

variables are grand mean centered in multi-level models.

26
I use Royston’s (2004) “ice” command in Stata to generate the imputed datasets and HLM6’s multiple imputation
capabilities to estimate the multiple models and combine results across models.

31
Results

Table 1 shows descriptive results on the degree to which Addhealth adolescents in

different types of neighborhoods agree that getting pregnant or getting someone pregnant would

“not be all that bad” at this time in their lives. As we would expect, adolescents in more

advantaged neighborhoods are more likely to strongly disagree that getting pregnant would not

be all that bad. However, what is more surprising from the perspective of social isolation theory

is that even in the most disadvantaged neighborhoods, over 70 percent of adolescents disagree or

strongly disagree that a pregnancy would not be all that bad. It would be very difficult to

conclude from Table 1 that the majority of adolescents in disadvantaged neighborhoods express

cultural models radically at odds with mainstream views on teenage pregnancy. Yet Table 1 also

reveals the greater heterogeneity of responses in disadvantaged neighborhoods. Descriptively,

what distinguishes disadvantaged from advantaged neighborhoods is the heterogeneity of

responses in disadvantaged neighborhoods. Whereas more advantaged neighborhoods have

considerable consensus on teenage pregnancy, in more disadvantaged neighborhoods there is a

sizable minority of adolescents who do not subscribe to the dominant view. Most research on

neighborhood context measures neighborhood characteristics using measures of central tendency

(such as the mean), but Table 1 reveals that internal variation (or what I am calling cultural

heterogeneity) also differs across neighborhoods, and that neighborhood mean and neighborhood

heterogeneity are empirically related.

Neighborhood Disadvantage and Cultural Heterogeneity

Table 2 examines the relationship between neighborhood disadvantage and cultural

heterogeneity controlling for other neighborhood characteristics that may be sources of

32
spuriousness. It examines Hypothesis 1, that disadvantaged neighborhoods will exhibit greater

cultural heterogeneity. Table 2 shows one OLS regression model each for neighborhood

heterogeneity in pregnancy frames and for neighborhood heterogeneity in relationship scripts.

The unit of analysis is the neighborhood (census tract). The key predictor is neighborhood

disadvantage. 27 Controls are included for percent Hispanic, racial diversity, percent foreign born,

and residential stability, though their coefficients are not of substantive interest for testing

Hypothesis 1. 28 As these control variables are highly correlated with one another, their individual

coefficients should not be interpreted as indicating their ability to predict cultural

heterogeneity. 29 Each model is weighted by the reliability of the neighborhood heterogeneity

measure. Descriptive statistics for the variables in these models are provided in Appendix B

Table B1.

Both models show a strong and statistically significant association between neighborhood

disadvantage and cultural heterogeneity, net of controls, and these associations are of the same

order of magnitude across domains. In the pregnancy frame model, a one standard deviation

increase in the neighborhood disadvantage scale is also associated with an increase of about one

fifth of a standard deviation in pregnancy frame heterogeneity. 30 In the relationship script model,

a one standard deviation increase in the neighborhood disadvantage scale is associated with an

27
Entering the neighborhood structural characteristics that make up the Neighborhood Disadvantage Scale in the
model individually produces similar results (not shown).
28
Percent Hispanic is given a quadratic functional form in these models because one would expect the association
between percent Hispanic and cultural heterogeneity to increase as the proportion of Hispanics increases from zero
but to peak and then decrease as the proportion Hispanic approaches one.
29
Multiple control variables are entered into the models individually so that they account for the maximum potential
omitted variable bias in the relationship between structural disadvantage and cultural heterogeneity. Drawing
conclusions about racial diversity, immigrant density, and residential stability as causes of cultural heterogeneity
would require a different model specification.
30
Recall that the neighborhood disadvantage variable is standardized because it is a scale with no inherent metric.
The standard deviation of the pregnancy frame heterogeneity variable is 0.3285 – see Appendix B Table B1.

33
increase of a little less than one sixth of a standard deviation in heterogeneity. These results

show that disadvantaged neighborhoods exhibit greater cultural heterogeneity. 31

Cultural Heterogeneity and Adolescent Outcomes

Hypothesis 2 is that adolescents in more culturally heterogeneous neighborhoods will be

less likely than those in more homogenous neighborhoods to act in accordance with their own

frames and scripts. This hypothesis predicts that the relationships between individual frames or

scripts and individual outcomes will be weaker in more heterogeneous neighborhoods. Table 3

shows various versions of the multi-level logit models predicting pre-marital sexual activity

described above. 32 (Descriptive statistics for the variables used in this model are provided in

Appendix B Table B2. Individual, family, and school characteristic control variable coefficients

are provided in Appendix B Table B3.) In model 1, the individual pregnancy frame is the

primary predictor. 33 This variable is interacted with dummy variables for quartiles of the

distribution of neighborhood heterogeneity in pregnancy frames. These quartile dummies are

also entered into the model directly. Converting the frame heterogeneity measure to quartiles

allows its effects to be nonlinear. The individual frame variable captures the relationship between

one’s pregnancy frame and pre-marital sexual behavior within low heterogeneity neighborhoods.

Not surprisingly, there is a strong relationship between an adolescent’s individual pregnancy

frame and his or her sexual behavior among adolescents in these neighborhoods. The interaction

31
One might question the relatively low R2 values produced by the models in Table 2. These are likely the result of
measurement error in the cultural heterogeneity measures due to the low reliability of measurement in
neighborhoods with few respondents. In OLS, measurement error in the dependent variable does not bias
coefficients but does increase error variance.
32
Cases were selected for the analysis in Table 3 if they had non-missing data on individual pregnancy frame and
neighborhood heterogeneity in pregnancy frames. Respondents under the age of 15 were not asked the survey
question measuring the pregnancy frame.
33
I experimented with entering this variable into the model as a series of dummy variables, but the relationship with
the outcome was fairly linear, so to reduce model complexity, I enter it as a linear term.

34
terms capture the difference in this relationship across neighborhoods with different levels of

frame heterogeneity. These coefficients show that as neighborhood frame heterogeneity

increases, the association between an adolescent’s own pregnancy frame and sexual behavior

declines dramatically.

Neighborhoods with greater heterogeneity in pregnancy frames also exhibit an average

frame that views teenage pregnancy less negatively (see Table 1). In other words, there is an

association between central tendancy and variation at the neighborhood level for pregnancy

frames. Thus, one potential source of spuriousness in Model 1 of Table 3 is a neighborhood’s

mean pregnancy frame. Model 2 adds the control for the neighborhood mean pregnancy frame.

Its coefficient is in the expected direction, but it is not statistically significant. Its inclusion

changes only slightly the primary coefficients of interest, the interaction terms. This model also

adds a term for the interaction between the individual pregnancy frame and the neighborhood

mean pregnancy frame. Its coefficient is small and statistically insignificant. These results

suggest that the smaller relationship between an adolescent’s own pregnancy frame and his or

her sexual behavior in more heterogeneous neighborhoods is not merely a product of greater

average permissiveness towards teenage pregnancy in such neighborhoods.

Moreover, the comparatively small magnitudes and lack of statistical significance of the

neighborhood mean pregnancy frame coefficients in predicting sexual activity suggest that once

cultural heterogeneity is taken into account, a neighborhood oppositional cultural environment is

not an important cause of differences in sexual behavior across neighborhoods. This is further

evidence against the social isolation view of the cultural context of disadvantaged

neighborhoods. Were social and/or cultural isolation affecting individual adolescent behavior,

35
one would expect neighborhood mean pregnancy frames to be strong predictors of individual

sexual activity.

It is also possible that the importance of neighborhood heterogeneity is a spurious result

of the strong association between neighborhood heterogeneity and neighborhood disadvantage.

Without controlling for neighborhood disadvantage, we cannot eliminate the possibility that the

effects of neighborhood heterogeneity are simply capturing the effects of neighborhood

disadvantage. Model 3 adds controls for neighborhood disadvantage that parallel the variables

measuring neighborhood heterogeneity. These include dummy variables for quartiles of the

distribution of neighborhood disadvantage and interactions between these quartiles and the

individual pregnancy frame. As we would expect, inclusion of the neighborhood disadvantage

variables reduces the magnitudes of the frame heterogeneity interaction coefficients, but they

remain large and statistically significant.

The coefficients in Model 3 show that in the lowest heterogeneity neighborhoods, the

relationship between individual frame and behavior is strong, but that the importance of the

respondent’s individual frame declines as neighborhood heterogeneity increases. In the lowest

heterogeneity neighborhoods, a one category increase in the individual pregnancy frame

multiples the odds of sexual behavior by about 0.61 (e-0.4975 = 0.6080), while in the highest

heterogeneity neighborhoods, a similar move in the individual pregnancy frame multiples the

odds of sexual behavior by a more modest 0.91 (e-0.4975+0.4059 = 0.9124). (Recall that since

exponentiated coefficients from logit models are odds ratios, values closer to 1 represent smaller

associations). Adolescents in higher heterogeneity neighborhoods are much less likely to act in

accordance with their own articulated frames regarding teenage pregnancy than their

counterparts in more culturally homogenous neighborhoods.

36
 One can also illustrate the interactions in these models graphically. Figure 1a shows the

relationship between predicted probabilities of sexual activity and individuals’ pregnancy frames

based on estimates from Model 3 in Table 3. Separate lines are plotted by quartiles of

neighborhood frame heterogeneity. Other variables are fixed at either zero (indicator variables)

or their means (continuous variables). Among adolescents living in neighborhoods with the least

cultural heterogeneity (those in the 1st and 2nd quartiles of heterogeneity), individual pregnancy

frames are strongly predictive of sexual activity. Among those in neighborhoods in the 3rd

quartile, there is still a negative sloping line, but the slope is considerably more modest. Among

those in neighborhoods with the highest pregnancy frame heterogeneity (those in the 4th

quartile), the relationship between one’s own pregnancy frame and one’s sexual behavior is

almost flat. Figure 1b rearranges the same data points to show the relationship between

neighborhood pregnancy frame heterogeneity and sexual activity, with separate lines plotted by

the adolescents’ own individual pregnancy frames. Here the importance of neighborhood cultural

heterogeneity is also evident. As neighborhood heterogeneity increases, the predicted

probabilities of sexual activity converge for all adolescents, no matter what their stated frame

regarding teenage pregnancy.

I now turn to the analysis of the romantic relationship scripts. Recall that the models are

slightly simpler here, as the failure to realize one’s ideal relationship script is measured directly.

Table 4 examines the relationship between neighborhood heterogeneity in ideal relationship

scripts and the difference between the ideal and actual relationship scripts of an individual. 34

(Descriptive statistics for the variables used in this model are provided in Appendix B Table B4.

Control variable coefficients are provided in Appendix B Table B5.) Model 1 shows that, net of

34
Cases were selected for analysis in Table 4 if they had non-missing data on the ideal relationship script and
neighborhood relationship script heterogeneity variables and had at least one romantic relationship that started after
the wave 1 in home interview (during which data on the ideal relationship script were collected).

37
individual, family, and school characteristics, neighborhood heterogeneity is a significant

predictor of the difference between one’s ideal and actual romantic relationships. Models 2 adds

controls for neighborhood disadvantage. This control reduces the heterogeneity coefficient, but it

remains statistically significant. These models provide further evidence in support of Hypothesis

2, suggesting that greater neighborhood cultural heterogeneity leads to a lower likelihood of

realizing one’s ideal relationship. To the extent that romantic relationship partners are selected

from among one’s neighbors, part of this relationship is likely related to the process of partner

selection. In more heterogeneous neighborhoods, it is more difficult to find a romantic partner

whose ideal relationship script is similar to one’s own. When two relationship scripts are

incompatible, at least one partner must fail to realize his or her ideal relationship if the

relationship continues. 35

Conclusion

This paper has attempted to reorient current thinking about the cultural context of

disadvantaged neighborhoods, particularly as it relates to adolescent outcomes. The canonical

account of neighborhood culture in the urban poverty literature (Wilson 1987) is one in which

the departure of middle class blacks from inner city neighborhoods has left behind a socially

isolated underclass with values that both depart significantly from middle class or mainstream

culture and are relatively uniform and coherent. Massey and Denton (1993) take this a step

further, writing that “ghetto culture has become an entity unto itself.” Theoretical notions such as

“oppositional culture” based on this premise now dominate much scholarship on inner city,

35
It is likely that there are important gender differences in adolescents’ ideal romantic relationship scripts.
Unfortunately, creating gender-specific measures of relationship script heterogeneity requires double the subject
density per neighborhood to maintain measurement reliability. It is not possible to create such gender-specific
measures in these data.

38
disadvantaged communities. Accordingly, in considering the impact of neighborhood culture on

adolescents, most previous work examining adolescent outcomes has employed a subculture

model, in which negative outcomes are the product of participation in a particular subculture

whose values promote negative behavior. Paradoxically, this work also emphasizes the multiple

and overlapping lifestyle groups or orientations in disadvantaged neighborhoods.

I have proposed an alternate view of the cultural context of disadvantaged

neighborhoods, a view that emphasizes their cultural diversity. The urban poverty literature has

lost track of the insights developed by Shaw and McKay (1969/1942) that poor communities are

characterized by conflicting cultural systems. Though Shaw and McKay were concerned with

crime and delinquency, these ideas have potential to further our understanding of the cultural

context of disadvantaged neighborhoods as it relates to other domains as well. I have argued that

an important way in which disadvantaged neighborhoods differ from more advantaged

neighborhoods is cultural heterogeneity, the presence of a wide array of competing and

contradictory cultural models.

The empirical analyses presented above show that cultural heterogeneity helps to predict

adolescent outcomes regarding sexual activity and romantic relationships when combined with

current theoretical concepts from cultural sociology, such as scripts, frames, and repertoires. This

paper has examined neighborhood cultural heterogeneity with regard to frames about teenage

pregnancy and scripts for romantic relationships. Using data from Addhealth, it has shown that

more disadvantaged neighborhoods exhibit greater cultural diversity in pregnancy frames and

relationship scripts. This paper also shows that adolescents in more culturally heterogeneous

neighborhoods are less likely to act in accordance with their own articulated scripts and frames.

In neighborhoods where there is greater diversity of frames regarding the consequences of

39
teenage pregnancy, adolescents are less likely to act in accordance with their own frames. In

neighborhoods where there is greater variation in scripts for romantic relationships, adolescents

are less likely to realize their own ideal relationships in the actual relationships that they pursue.

From a theoretical perspective, this research has implications for both social isolation and

social organization theories of neighborhood effects. The importance of cultural heterogeneity

calls into question one of the fundamental ideas of social isolation theory, that disadvantaged

neighborhoods are isolated from mainstream or middle class culture. Instead, these results

suggest that disadvantaged neighborhoods contain a mix of cultural models and that this

heterogeneity impacts neighborhood adolescents. In contrast, the findings in this study are

generally supportive of neighborhood effects models based on social organization theory and

move us toward understanding the mechanisms underlying these models. Social disorganization

theory highlights the limited capacity of residents in disadvantaged communities to regulate the

behaviors of their neighbors. Heterogeneity in cultural lifestyles or orientations can be

understood as the failure of more middle class or mainstream residents of disadvantaged

neighborhoods to regulate behavior in their communities. The result of this failure is that neither

“oppositional” nor “mainstream” behavior is dominant in disadvantaged communities.

More broadly, this paper’s findings challenge the underclass concept as a fruitful

framework for the analysis of life in poor urban neighborhoods. This study suggests that rather

than being disconnected from mainstream society, poor urban residents live in cultural

environments that include both local elements and elements from the wider society and that these

disparate elements are often contradictory and competing. This means that, as has been described

in the ethnographic literature (Anderson 1990, 1999, Newman 1999), cultural conflict and

cultural disagreement are widespread. I have argued here that such conflict may have further

40
consequences for wellbeing and decision-making. In addition, a theoretical framework based on

cultural heterogeneity allows both for variation in individual outcomes among those who

experience the same structural conditions and for greater agency, as individuals choose between

and evaluate the cultural models that exist in their cultural environments and in their repertoires.

The cultural heterogeneity view is both more optimistic and in some ways more

pessimistic than the underclass view. On the one hand, since according to the cultural

heterogeneity view there is widespread support for conventional cultural models, the seeds of

positive change are likely present in the vast majority of disadvantaged neighborhoods. On the

other hand, neighborhood cultural heterogeneity can be created when even a relatively small

minority of individuals or families display alternative or oppositional behaviors or lifestyles and

thereby disrupt, confuse, or weaken the dominance of conventional cultural models. Cultural

heterogeneity, and its consequences, will likely persist until the most disadvantaged residents of

poor neighborhoods are empowered with more positive ways to adapt to and rise above poverty.

Though this study has not set out to explicitly test competing hypotheses regarding

cultural theories, it does suggest some implications for cultural sociology. First, this paper’s

findings regarding the consequences of neighborhood cultural heterogeneity show how local or

group cultural contexts can matter for behavior, even when those local cultures are not starkly

distinct or separate from other local cultures or the wider culture. In other words, different

behavioral outcomes do not require distinct subcultures. Second, this paper provides another

example of the utility of cognitive cultural concepts such as frames, scripts, and repertoires. In

doing so, it also suggests a further specification of how repertoires operate. The activation of

elements from one’s repertoire depends not just on one’s structural position (Lamont 1992) but

also on the relationships between various elements in the repertoire and on local social support

41
for those elements. Third, the importance of cultural heterogeneity suggests that not only are the

meanings attached to cultural models important, but how those meanings are organized and

related to one another also has implications for behavior. The degree of harmony or conflict

among available cultural models is another dimension along which cultural contexts or cultural

repertoires can be evaluated.

While the results in this study are strongly suggestive of the importance of cultural

heterogeneity for understanding the cultural context of disadvantaged neighborhoods, further

research is required. Better measures of frames and scripts would provide stronger tests of the

hypotheses examined here. For example, the measure of heterogeneity of pregnancy frames does

not capture the presence of multiple frames but rather variation in allegiance to a particular

frame. 36 Further empirical predictions can be generated to test the utility of cultural

heterogeneity concept. For example, the hypothesized weaker commitment to frames or scripts

that comes from exposure to greater cultural heterogeneity would suggest that adolescents in

more heterogeneous neighborhoods should exhibit greater temporal variation in their articulated

frames and scripts. Finally, further research is needed to examine the degree of cultural

heterogeneity in other domains such as education or health, the ability of cultural heterogeneity

to predict other outcomes, the importance of cultural heterogeneity in other contexts such as

schools, and whether there are gender contingent effects of cultural heterogeneity.

36
In addition, it could be argued that the survey question used here to measure pregnancy frames also taps the moral
status of teenage pregnancy because it contains the phrase “all that bad,” or it could be argued that the question
wording is confusing to adolescents because it contains a double negative. Nevertheless, this measure is strongly
predictive of behavior among adolescents in culturally homogenous neighborhoods, so in order to call into question
the larger argument, measurement problems would have to disproportionately affect adolescents in culturally
heterogeneous neighborhoods. Still, whether or not a moral element is also captured must be determined by future
research. More generally, social scientists have spent decades and millions of dollars developing and fine-tuning
survey measures of structural position, but we have only begun to develop valid and reliable survey measures of
cultural concepts.

42
References

Abbott, Andrew. 1990. “A Primer on Sequence Methods.” Organization Science 1: 375-392.

Abbott, Andrew. 1995. “Sequence Analysis: New Methods for Old Ideas.” Annual Review of
Sociology 21: 93-113.

Abbott, Andrew and Alexandra Hrycak. 1990. “Measuring Resemblance in Sequence Data: An
Optimal Matching Analysis of Musicians’ Careers.” American Journal of Sociology 96:
144-85.

Acock, Alan C. 2005. “Working With Missing Values.” Journal of Marriage and Family 67:
1012-1028.

Ainsworth-Darnell, James and Douglas Downey. 1998. “Assessing the Oppositional Culture
Explanation for Racial/Ethnic Differences in School Performance.” American
Sociological Review. 63:536-553.

Ajzen, Icek. 2001. “Nature and Operation of Attitudes.” Annual Review of Psychology 52: 27-58.

Allison, Paul D. 2002. Missing Data. Thousand Oaks, CA: Sage.

Anderson, Elijah. 1978. A Place on the Corner. Chicago: University of Chicago Press.

Anderson, Elijah. 1990. Streetwise: Race, Class, and Change in an Urban Community.
Chicago: University of Chicago Press.

Anderson, Elijah. 1991. “Neighborhood Effects on Teenage Pregnancy.” p. 375-398 in

Christopher Jencks and Paul E. Peterson, eds. The Urban Underclass. Washington, DC:
Brookings Institution.

Anderson, Elijah. 1999. Code of the Street: Decency, Violence, and the Moral Life of the Inner
City. New York: W. W. Norton.

Barber, Jennifer S. 2001. “Ideational Influences on the Transition to Parenthood: Attitudes

Toward Childbearing and Competing Alternatives.” Social Psychology Quarterly 64:101-
127.

Benford, Robert D. and David A. Snow. 2000. “Framing Processes and Social Movements: An
Overview and Assessment.” Annual Review of Sociology 26: 611-639.

Blair, Julian and Michael G. Lacy. 2000. “Statistics of Ordinal Variation.” Sociological Methods
and Research 28: 251-279.

Bourgois, Phillipe. 1995. In Search of Respect: Selling Crack in El Barrio. Cambridge:

Cambridge University Press.

43
Bourdieu, Pierre. 1984 [1979]. Distinction: A Social Critique of the Judgment of Taste.
Cambridge, MA: Harvard University Press.

Bourdieu, Pierre and Jean-Claude Passeron. 1977 [1970]. Reproduction in Education, Society
and Culture. Beverly Hills: Sage.

Brooks-Gunn, Jeanne, Greg J. Duncan, Pamela K. Klebanov, and Naomi Sealand. 1993. “Do
Neighborhoods Influence Child and Adolescent Development?” American Journal of
Sociology 99:353–95.

Browning, Christopher R., Tama Leventhal, and Jeanne Brooks-Gunn. 2004. “Neighborhood
Context and Racial Differences in Early Adolescent Sexual Activity.” Demography 41:
697-720.

Browning, Chrisopher R., Tama Leventhal, and Jeanne Brooks-Gunn. 2005. “Sexual Initiation in
Early Adolescence: The Nexus of Parental and Community Control.” American
Sociological Review 70: 758-778.

Bryson, Bethany. 1996. "Anything But Heavy Metal: Symbolic Exclusion and Musical
Dislikes." American Sociological Review 61(5): 884-99.

Bursik, Robert J., Jr. and Harold G. Grasmick. 1993. Neighborhoods and Crime: The
Dimensions of Effective Community Control. New York: Lexington Books.

Carter, Prudence. 2005. Keepin' It Real: School Success Beyond Black and White. New York:
Oxford University Press.

Cook, Phillip and Jens Ludwig. 1998. “The Burden of ‘Acting White’: Do Black Adolescents
Disparage Academic Achievement?” Pp. 375-401 in Christopher Jencks and Meredith
Phillips (eds) The Black White Test Score Gap. Washington, D.C.: Brookings.

DiMaggio, Paul. 1982. "Cultural Capital and School Success: the Impact of Status Culture
Participation of the Grades of U.S. High School Students." American Sociological
Review 47: 189-201.

DiMaggio, Paul. 1997. “Culture and Cognition.” Annual Review of Sociology 23: 263-87.

Dohan, Daniel. 2003. The Price of Poverty: Money, Work, and Culture in the Mexican American
Barrio. Berkeley: University of California Press.

Dubois, W. E. B. 1996 [1899]. The Philadelphia Negro: A Social Study. Philadelphia: University
of Pennsylvania Press.

Duneier, Mitchell. 1992. Slim’s Table: Race, Respectability, and Masculinity. Chicago:
University of Chicago Press.

44
Edin, Kathryn and Maria Kefalas. 2005. Promises I Can Keep: Why Poor Women Put
Motherhood Before Marriage. Berkeley: University of California Press.

Erikson, Bonnie. 1996. "Culture, Class and Connections." American Journal of Sociology 102:
217-51.

Fernandez. Roberto M. and David Harris. 1992. “Social Isolation and the Underclass.” p. 257-
293 in Adele V. Harrell and George E. Peterson, eds. Drugs, Crime, and Social Isolation:
Barriers to Urban Opportunity. Washington, DC: Urban Institute Press.Frazier, E.
Franklin. 1966. The Negro Family in the United States. Revised and Abridged edition.
Chicago: University of Chicago Press.

Fine, Gary Alan. 1979. “Small Groups and Culture Creation: The Idioculture of Little League
Baseball Teams.” American Sociological Review 44: 733-745.

Fine, Gary Alan and Sherryl Kleinman. 1979. “Rethinking Subculture: An Interactionist
Analysis.” American Journal of Sociology 85: 1-20.

Fishbein, Martin and Icek Ajzen. 1975. Belief, Attitude, Intention, and Behavior: An Introduction
to Theory and Research. Reading, MA: Addison-Wesley.

Fordham, Signithia and John Ogbu. 1986. “Black Students’ School Success: Coping with the
Burden of ‘Acting White.’” Urban Review 18(3):176-206.

Fuller, Bruce, Susan D. Holloway, Marylee F. Rambaud, and Constanza Eggers-Pierola. 1996.
“How Do Mothers Choose Child Care? Alternative Cultural Models in Poor
Neighborhoods.” Sociology of Education 69: 83-104.

Furstenberg, Frank F. et al (eds). 1999. Managing to Make It: Urban Families and Adolescent
Success. Chicago: University of Chicago Press.

Gans, Herbert. 1962. Urban Villagers. New York: Free Press.

Goldenberg, Claude, Ronald Gallimore, Leslie Reese, and Helen Garnier. 2001. “Cause or
Effect? A Longitudinal Study of Immigrant Latino parents’ Aspirations and Expectations,
and Their Children’s School Performance.” American Educational Research Journal
38(3): 547-582.

Goffman, Erving. 1959. The Presentation of Self in Everyday Life. New York: Doubleday.

Goffman, Erving. 1974. Frame Analysis. Cambridge: Harvard University Press.

Hannerz, Ulf. 1969. Soulside: Inquiries into Ghetto Culture and Community. New York:
Columbia University Press.

45
Harding, David J. 2003. "Counterfactual Models of Neighborhood Effects: The Effect of
Neighborhood Poverty on Dropping Out and Teenage Pregnancy." American Journal of
Sociology 109(3): 676-719.

Harris, Kathleen Mullan, Francesca Florey, Joyce Tabor, Peter S. Bearman, Jo Jones, and J.
Richard Udry. 2003 The National Longitudinal Study of Adolescent Health: Research
Design [WWW document]. URL: http://www.cpc.unc.edu/projects/addhealth/design.

Hayes, Sharon. 2003. Flat Broke With Children: Women in the Age of Welfare Reform. New
York: Oxford University Press.

Hochschild, Jennifer L. 1995. Facing Up to the American Dream: Race, Class, and the Soul of
the Nation. Princeton, NJ: Princeton University Press.

Holloway, Susan D., Bruce Fuller, Marylee F. Rambaud, and Constanza Eggers-Pierola. 1997.
Through My Own Eyes: Single Mothers and the Cultures of Poverty. Cambridge, MA:
Harvard University Press.

Jackson, John L. 2001. Harlemworld: Doing Race and Class in Contemporary Black America.
Chicago: University of Chicago Press.

Kefalas, Maria. 2003. Working Class Heroes: Protecting Home, Community, and Nation in a
Chicago Neighborhood. Berkeley: University of California Press.

Kornhauser, Ruth R. 1978. Social Sources of Delinquency. Chicago: University of Chicago Press.

Lamont, Michèle. 1992. Money, Morals and Manners: The Culture of the French and the
American Upper-Middle Class. Chicago: University of Chicago Press.

Lamont, Michèle. 2000. The Dignity of Working Men: Morality and the Boundaries of Race,
Class, and Immigration. Cambridge, MA: Harvard University Press.

Lamont, Michèle and Annette Lareau 1988. "Cultural Capital: Allusions, Gaps, and Glissandos
in Recent Theoretical Developments." Sociological Theory 6(2): 153-68.

Lamont, Michèle and Laurent Thévenot. 2000. Rethinking Comparative Cultural Sociology:
Repertoires of Evaluation in France and the United States. Cambridge: Cambridge
University Press.

Lamont, Michèle, and Viràg Molnár. 2002. "The Study of Boundaries Across the Social
Sciences." Annual Review of Sociology 28: 167-195.

Lareau, Annette. 2004. Unequal Childhoods: Class, Race, and Family Life. Berkeley, CA:
University of California.

46
Lareau, Annette. 2000. Home Advantage: Social Class and Parental Intervention in
Elementary Education, Second Edition, Lanham, MD, Rowan and Littlefield.

Lewis, Oscar. 1966. La Vida: A Puerto Rican Family in the Culture of Poverty – San Juan and
New York. New York: Random House.

Lewis, Oscar. 1969. “The Culture of Poverty,” in Daniel Patrick Moynihan, ed. On
Understanding Poverty: Perspectives from the Social Sciences. New York: Basic Books.

Liebow, Elliot. 1967. Tally’s Corner: A Study of Negro Streetcorner Men. Boston: Little,
Brown.

Liska, Allen E. 1984. “A Critical Examination of the Causal Structure of the Fishbein/Ajzen
Attitude Behavior Model.” Social Psychology Quarterly 47: 61-74.

Little, Roderick J. A. and Donald B. Rubin. 2002. Statistical Analysis With Missing Data, 2nd
Edition. New York: Wiley.

Massey, Douglas and Nancy Denton. 1993. American Apartheid: Segregation and the Making of
an Underclass. Cambridge, MA: Harvard University Press.

MacLeod, Jay. 1995. Ain’t No Makin’ It: Aspirations and Attainment in a Low-Income
Neighborhood. Boulder: Westview.

Mickelson, Roslyn A. 1990. “The Attitude-Achievement Paradox Among Black Adolescents.”

Sociology of Education 63: 44-61.

Moynihan, Daniel P. 1965. The Negro Family: The Case for National Action. Washington, DC:
US Department of Labor.

Newman, Katherine S. 1992. “Culture and Structure in The Truly Disadvantaged.” City and
Society 6(1): 3-25.

Newman, Katherine S. 1999. No Shame in My Game: The Working Poor in the Inner City. New
York: Vintage and Russell Sage.

Newman, Katherine S. 2003. A Different Shade of Gray: Mid-Life and Beyond in the Inner City.
New York: New Press.

Newman, Katherine S. 2006. Chutes and Ladders: Navigating the Low-wage Labor Market.
New York: Harvard University Press and Russell Sage.

Ogbu, John U. 2004. “Collective Identity and the Burden of ‘Acting White’ in Black History.
Community, and Education.” The Urban Review 36: 1-35.

Park, Robert E. and Ernest W. Burgess. 1925. The City. Chicago: University of Chicago Press.

47
Parkin, Frank. 1971. Class Inequality and Political Order: Social Stratification in Capitalist and
Communist Societies. London: MacGibbon and Lee.

Patillo-McCoy, Mary. 1998. “Church Culture as a Strategy of Action in the Black Community,”
American Sociological Review 63: 767-784.

Pattillo-McCoy, Mary. 1999. Black Picket Fences: Privilege and Peril Among the Black Middle
Class. Chicago: University of Chicago Press.

Quinn, Naomi and Dorothy Holland. 1987. “Culture and Cognition.” p. 3-40 in Dorothy Holland
and Naomi Quinn, eds. Cultural Models in Language and Thought. Cambridge:
Cambridge University Press.

Rankin, Bruce H. and James M. Quane. 2000. “Neighborhood Poverty and the Social Isolation of
Inner-City African American Families.” Social Forces 79(1): 139-164.

Reardon, Sean F. and Glenn Firebaugh. 2002. “Measures of Multigroup Segregation.”

Sociological Methodology 32: 33-67.

Rodman, Hyman. 1963. “The Lower-Class Value Stretch.” Social Forces 42: 205-215.

Rohwer, Goetz and Ulrich Poetter. 2005. Transition Data Analysis (version 6.4). Computer
software. Available from http://steinhaus.stat.ruhr-uni-bochum.de/tda.html.

Royston, Patrick. 2004. “Multiple imputation of missing values.” The Stata Journal 4: 227-241.

Sankoff, David and Joseph Kruskal. 1983. Time Warps, String Edits, and Macromolecules: The
Theory and Practice of Sequence Comparison. Reading, MA: Addison-Wesley.

Sampson, Robert, Stephen Raudenbush, and Felton Earls. 1997. “Neighborhoods and Violent
Crime: A Multilevel Study of Collective Efficacy.” Science. 227:918-924.

Sayer, Liana C., Nathan Wright, and Kathryn Edin. 2005. “Differences in Family Attitudes by
Education.” Revision of paper presented at the annual meetings of the Population
Association of America, Minneapolis, Minnesota, May 2003.

Schudson, Michael. 1989. “How Culture Works: Perspectives from media studies on the efficacy
of symbols.” Theory and Society 18: 153-180.

Sewell, William. 1992. “A Theory of Structure: Duality, Agency, and Transformation.”

American Journal of Sociology 98: 1-29.

Sewell, William. 1999. “The Concept(s) of Culture.” Pp. 35-61 in Victoria E. Bonnell and Lynn
Hunt, eds. Beyond the Cultural Turn. Berkeley: Univ. of California Press.

48
Shaw, Clifford R. 1929. Delinquency Areas. Chicago: University of Chicago Press.

Shaw, Clifford R. and Henry D. McKay. 1969/1942. Juvenile Delinquency and Urban Areas,
Revised Edition. Chicago: University of Chicago Press.

Small, Mario L. 2002. “Culture, Cohorts, and Social Organization Theory: Understanding Local
Participation in a Latino Housing Project.” American Journal of Sociology 108:1-54.

Small, Mario L. 2004. Villa Victoria: The Transformation of Social Capital in a Boston Barrio.
Chicago: University of Chicago Press.

Solorzano, Daniel G. 1992. “An Exploratory Analysis of the Effects of Race, Class and Gender
on Student and Parent Mobility Aspirations.” Journal of Negro Education 61(1): 30-44.

Stoval, Katherine, Michael Savage, and Peter Bearman. 1996. “Ascription into Achievement:
Models of Career Systems at Lloyds Bank, 1890-1970.” American Journal of Sociology
102: 358-399.

Sullivan, Mercer. 1989. Getting Paid: Youth, Crime, and Work in the Inner City. Ithaca: Cornell
University Press.

Suttles, Gerald. 1968. The Social Order of the Slum: Ethnicity and Territory in the Inner City.
Chicago: University of Chicago Press.

Swidler, Ann. 1986. “Culture in Action: Symbols and Strategies.” American Sociological
Review 51:273-86.

Swidler, Ann. 2001. Talk of Love. Chicago: University of Chicago Press.

Teitler, Julien, and Christopher Weiss. 2000. “Effects of neighborhood and school environments
on transitions to first sexual intercourse.” Sociology of Education 73: 112-132.

Tigges, Leann M., Irene Browne, and Gary P. Green. 1998. “Social Isolation of the Urban Poor:
Race, Class, and Neighborhood Effects in Social Resources.” Sociological Quarterly
39(1): 53-77.

Tilly, Charles. 1978. From Mobilization to Revolution. Reading, MA: Addison-Wesley.

Valentine, Charles A. 1968. Culture and Poverty: Critique and Counter-Proposals. Chicago:
University of Chicago Press.

Venkatesh, Sudhir A. 2000. American Project. Chicago: University of Chicago Press.

Waller, Maureen R. 2002. My Baby’s Father: Unmarried Parents and Paternal Responsibility.
Ithaca: Cornell University Press.

49
White, Michael J. 1986. “Segregation and Diversity Measures in Population Distribution.”
Population Index 52: 198-221.

Willis, Paul. 1977. Learning to Labor. New York: Columbia University Press.

Wilson, William Julius. 1987. The Truly Disadvantaged: The Inner-City, the Underclass, and
Public Policy. Chicago: University of Chicago Press.

Wilson, William Julius. 1996. When Work Disappears: The World of the New Urban Poor.
New York: Knopf.

Young, Alford A, Jr. 2004. The Minds of Marginalized Black Men: Making Sense of Mobility,
Opportunity, and Life Chances. Princeton: Princeton University Press.

50
Table 1: Pregnancy Frame by Neighborhood Disadvantage
“It wouldn’t be all that bad if you got pregnant [or got someone pregnant] at this time in your life.”
Neighborhood Neither
Disadvantage Strongly Agree Nor Strongly
Scale Quintiles Agree Agree Disagree Disagree Disagree Row Obs
1st 1.1% 4.0% 5.2% 31.7% 58.1% 1,899
2nd 1.6% 5.3% 7.9% 32.1% 53.1% 3,508
3rd 1.2% 6.8% 9.3% 31.7% 50.9% 3,880
4th 2.9% 9.7% 10.1% 32.6% 44.8% 2,689
5th 4.0% 12.4% 11.1% 35.2% 37.4% 1,640

All 2.0% 7.4% 8.7% 32.4% 49.5% 13,616

Note: Estimates account for Addhealth sample design and Addhealth In-Home Wave 1 weight

51
Table 2: OLS Models of Neighborhood Cultural Heterogeneity
Model 1 Model 2
Pregnancy Relationship
Frame Script
Nhood Disadvantage Scale 0.06786* 0.15083*
(0.00872) (0.02704)

% Hispanic -0.00014 -0.00069

(0.00125) (0.00398)

% Hispanic Squared 0.00001 -0.00002

(0.00014) (0.00004)

Racial Diversity Index 0.00102* 0.00102

(0.00043) (0.00138)

% Foreign Born 0.00074 0.00429

(0.00080) (0.00264)

% Owner Occupied 0.00026 -0.00283*

(0.00041) (0.00133)

% Units Occupied 5 Years 0.00031 -0.00087

(0.00055) (0.00178)

Constant 0.56636* 0.14836

(0.04751) (0.15386)

N neighborhoods 1322 1367

R2 0.086 0.057
NOTES:
Each model weighted by reliability of neighborhood heterogeneity measure
Descriptive statistics available in Appendix B Table B1

52
Table 3: Three-level Logit Models Predicting Sex Between Waves I and II
Model 1 Model 2 Model 3
Individual Pregnancy Frame -0.4384* -0.4185* -0.4975*
(0.0873) (0.0988) (0.1030)

Nhood Mean Pregnancy Frame -0.1504 -0.0930

(0.1128) (0.1123)

Individual Frame X Nhood Mean Pregnancy Frame -0.0196 0.0477

(0.0830) (0.0868)

Pregnancy Frame Nhood Heterogeneity Quartiles:

2nd Quartile vs. 1st Quartile 0.0595 0.0200 0.0351
(0.1006) (0.1104) (0.1115)
3rd Quartile vs. 1st Quartile 0.0517 -0.0360 -0.0368
(0.1019) (0.1285) (0.1289)
Top Quartile vs. 1st Quartile 0.1503 0.0144 0.0159
(0.1648) (0.1986) (0.1947)

Individual Frame X NHood Frame Heterogeneity Quartiles:

2nd Quartile vs. 1st Quartile 0.2100* 0.1949 0.1727
(0.1001) (0.1055) (0.1040)
3rd Quartile vs. 1st Quartile 0.3745* 0.3514* 0.3144*
(0.0902) (0.1161) (0.1189)
Top Quartile vs. 1st Quartile 0.4335* 0.4163* 0.4059*
(0.1381) (0.1560) (0.1654)
Nhood Disadvantage Scale Quartiles:
2nd Quartile vs. 1st Quartile 0.0965
(0.0767)
3rd Quartile vs. 1st Quartile 0.1408
(0.0881)
Top Quartile vs. 1st Quartile 0.2160*
(0.1057)
Individual Frame X Nhood Disadvantage Scale Quartiles:
2nd Quartile vs. 1st Quartile 0.1361
(0.0729)
3rd Quartile vs. 1st Quartile 0.0783
(0.0765)
Top Quartile vs. 1st Quartile 0.2419*
(0.0809)

Constant -0.5372* -0.4742* -0.5926*

(0.1521) (0.1714) (0.1824)

N individuals 9,281 9,281 9,281

N neighborhoods 1,357 1,357 1,357
N schools 142 142 142
NOTES:
Models weighted by reliability of neighborhood frame means
Robust standard errors in parentheses
Descriptive statistics available in Appendix B Table B2
Individual, family, and school control variable coefficients available in Appendix B Table B3
Missing values on control variables imputed using multiple imputation with 10 replications

53
Table 4: Three-Level Models Predicting Difference Between Ideal Relationship and Actual Relationship Scripts
Model 1 Model 2
Nhood Heterogeneity Ideal Relationship Script 0.0539* 0.0473*
(0.0154) (0.0156)

Nhood Disadvantage Scale 0.0237

(0.0128)

Constant 2.0518* 2.0509*

(0.0310) (0.0310)

N individuals 5,916 5,916

N neighborhoods 1,299 1,299
N schools 143 143
NOTES:
Models weighted by reliability of neighborhood script heterogeneity means
Robust standard errors in parentheses
Descriptive statistics available in Appendix B Table B4
Individual, family, and school control variable coefficients available in Appendix B Table B5
Missing values on control variables imputed using multiple imputation with 10 replications

54
 Figure 1a: Relationship Between Individual Pregnancy Frame
and Sexual Activity
(by Neighborhood Pregnancy Frame Heterogeneity Quartiles)
0.8

1st Quartile
0.7
2nd Quartile
0.6
Probability of Sexual Activity

3rd Quartile
0.5

0.4 4th Quartile

0.3

0.2

0.1

0
Strongly Agree Agree Neither Disagree Strongly Disagree
Individual Pregnancy Frame

55
 Figure 1b: Relationship Between Neighborhood Pregnancy Frame
Heterogeneity and Sexual Activity
(by Individual Pregnancy Frame)
0.8

0.7
Probability of Sexual Activity

0.6 Strongly Agree

Agree

0.5
Neither
0.4
Disagree
0.3

0.2
Strongly Disagree
0.1

0
1 2 3 4
Heterogeneity Quartiles

56
 Appendix A: Descriptions of Individual, Family, and School Control Variables

(All measured at Wave 1)

Individual Characteristics:

Race/Ethnicity: A set of indicator (0/1) variables for the adolescent’s race and ethnicity. In

Addhealth, the adolescent can self identify as belonging to one or more categories, including

White, Black, Native American, Asian, or other Race. White is the omitted category in models. I

also include an indicator variable for those adolescents who choose more than one category. The

adolescent can also choose to identify as Hispanic/Latino or not.

Immigrant: An indicator for those born outside the United States.

Low Birth Weight: An indicator for weighing less than 88 ounces (5.5 lbs.) at birth.

Mother’s Age at Birth: The age in years of the mother when the adolescent was born.

Family Characteristics:

Home Language not English: An indicator variable for those adolescents whose family does not

speak English at home.

Household Size: The number of persons living in the adolescent’s household.

Household Type: A set of indicator variables for the family type: Married, Single Parent, and

Other (which includes step parent families). Married is the omitted category.

Parent variables are based on the primary residential parent who completed the parent

questionnaire, usually the biological mother but sometimes the father or other caretaker.

Parent Immigrant: Primary parent not born in the US.

57
Parent Education: A set of indicator variables for the primary parent’s completed level of

education: less than High School, High School Graduate, Some College or Trade School, and

College Graduate. Less than high school is the omitted category.

Parent Professional Occupation: Primary parent currently works in a managerial or professional

occupation.

Parent Disabled: Primary parent is mentally or physically handicapped.

Parent Welfare Receipt: Primary parent currently receives welfare, either for self or for the

adolescent.

Log Family Income: The natural logarithm of the household’s total income in thousands of

dollars, as reported in the parent questionnaire.

School Characteristics:

Urbanicity: A set of indicator variables for the location of the school: Urban, Suburban, or Rural.

Suburban is the omitted category.

School Size: A set of indicator variables for the number of students at the school, Small (< 400),

Medium (400-1000), and Large (> 1000). Medium is the omitted category.

Cumulative Dropout Rate: The proportion of students who begin the school in its lowest grade

who complete its highest grade.

Percent College Prep Program: The proportion of 12th graders who are enrolled in an academic

or college prep program.

Catholic School: An indicator for Catholic schools.

Private School: An indicator for all other non-public schools.

58
 Appendix B: Supplementary Tables

Table B1: Descriptive Statistics for Variables in Table 2

Pregnancy Frame Model (n=1322) Mean SD Min Max
Pregnancy Frame Heterogeneity 0.5338 0.3285 0 1
Nhood Disadvantage Scale 0 1 -3.0494 5.4209
% Hispanic 13.6931 22.0561 0 96.2677
% Hispanic Squared 673.6044 1750.7110 0 9267.4630
Racial Diversity Index 26.8420 20.8092 0 74.3581
% Foreign Born 11.3441 16.3284 0 86.9001
% Owner Occupied 59.6721 22.7260 0 97.8407
% Units Occupied 5 Years 48.2259 13.8744 9.7990 97.5802
Reliability (weight) 0.2523 0.2153 0.0559 0.9345

Relationship Script Model (n=1367) Mean SD Min Max

Relationship Script Heterogeneity 0 1 -2.8475 6.3506
Nhood Disadvantage Scale 0 1 -2.9985 5.4274
% Hispanic 13.4554 21.9078 0 96.2677
% Hispanic Squared 660.6486 1735.4820 0 9267.4630
Racial Diversity Index 26.6161 20.7642 0 74.3581
% Foreign Born 11.1963 16.1379 0 86.9001
% Owner Occupied 60.1739 22.5801 0 97.8407
% Units Occupied 5 Years 48.2388 13.9728 9.7990 100.0000
Reliability (weight) 0.5076 0.3476 0.1090 0.9998

59
Table B2: Descriptive Statistics for Variables in Table 3
Mean SD Min Max
Individual/Family Level Variables (n=9,281)
Sexual Activity Between Waves 1 and 2 0.43 0.49 0 1
Pregnancy Frame 4.21 1.00 1 5
Female 0.50 0.50 0 1
Age 16.25 1.04 15 21
Hispanic 0.18 0.39 0 1
Black 0.21 0.41 0 1
Native American 0.04 0.18 0 1
Asian 0.08 0.28 0 1
Other Race 0.10 0.30 0 1
Multi Race 0.05 0.21 0 1
Home Language Not English 0.13 0.34 0 1
Immigrant 0.10 0.30 0 1
Household Size 4.72 1.68 1 21
Single Parent Household 0.24 0.43 0 1
Other Household Type 0.23 0.42 0 1
Parent Immigrant 0.22 0.41 0 1
Parent Education – HS Grad 0.29 0.45 0 1
Parent Education – Some College 0.27 0.44 0 1
Parent Education -- College 0.23 0.42 0 1
Parent Professional/Managerial Occ 0.32 0.47 0 1
Parent Disabled 0.05 0.22 0 1
Family Welfare Receipt 0.11 0.31 0 1
Log Family Income 3.50 0.87 0 6.91
Low Birth Weight 0.11 0.32 0 1
Mother’s Age at Birth 25.5 5.40 5 53

Neighborhood Level Variables (n=1,357)

Neighborhood Disadvantage Scale 0 0.70 -2.22 3.91
Pregnancy Frame Nhood Heterogeneity 0.49 0.30 0 1
Pregnancy Frame Nhood Mean 4.19 0.61 1 5
Nhood Reliability (weight) 0.30 0.23 0.06 0.93

School Level Variables (n=142)

Urban 0.30 0.46 0 1
Rural 0.13 0.34 0 1
Small 0.21 0.41 0 1
Large 0.31 0.46 0 1
Cumulative Dropout Rate 7.80 11.2 0 68.52
Percent in College Prep Program 40.44 34.15 0 100
Catholic School 0.04 0.20 0 1
Private School 0.02 0.14 0 1

60
Table B3: Control Variable Coefficients for Models in Table 3
Model 1 Model 2 Model 3
Individual/Family Level Variables
Female 0.1697* 0.1702* 0.1682*
(0.0554) (0.0556) (0.0561)
Age 0.3007* 0.3001* 0.3004*
(0.0230) (0.0231) (0.0232)
Hispanic 0.1040 0.0962 0.0947
(0.0995) (0.0993) (0.0990)
Black 0.1617* 0.1498 0.1392
(0.0816) (0.0825) (0.0823)
Native American -0.0662 -0.0735 -0.0836
(0.1144) (0.1138) (0.1141)
Asian -0.3709* -0.3787* -0.3768*
(0.1102) (0.1118) (0.1102)
Other Race 0.0636 0.0576 0.0553
(0.0762) (0.0776) (0.0766)
Multi Race 0.2394* 0.2479* 0.2539*
(0.1169) (0.1173) (0.1170)
Home Language Not English -0.3728* -0.3794* -0.3922*
(0.1130) (0.1144) (0.1137)
Immigrant -0.4069* -0.4113* -0.4065*
(0.0625) (0.0621) (0.0622)
Household Size -0.0351 -0.0365 -0.0373
(0.0151) (0.0152) (0.0152)
Single Parent Household 0.3649* 0.3535* 0.3568*
(0.0903) (0.0921) (0.0918)
Other Household Type 0.3977* 0.3923* 0.3921*
(0.0743) (0.0761) (0.0776)
Parent Immigrant -0.0241 -0.0193 -0.0166
(0.0822) (0.0825) (0.0824)
Parent Education – HS Grad 0.1368 0.1346 0.1323
(0.0943) (0.0935) (0.0931)
Parent Education – Some College 0.1233 0.1258 0.1221
(0.0901) (0.0902) (0.0900)
Parent Education -- College 0.0550 0.0651 0.0680
(0.1073) (0.1077) (0.1079)
Parent Professional/Managerial Occ -0.0482 -0.0499 -0.0444
(0.0636) (0.0627) (0.0628)
Parent Disabled -0.1206 -0.1298 -0.1337
(0.1017) (0.0960) (0.0951)
Family Welfare Receipt 0.0531 0.0507 0.0471
(0.0735) (0.0740) (0.0733)
Log Family Income -0.0168 -0.0299 -0.0252
(0.0437) (0.0394) (0.0395)
Low Birth Weight -0.1944* -0.1788* -0.1805*
(0.0848) (0.0819) (0.0817)
Mother’s Age at Birth -0.0173* -0.0185* -0.0182*
(0.0063) (0.0067) (0.0066)
School Level Variables
Urban -0.0968 -0.0992 -0.1134
(0.0865) (0.0870) (0.0878)
Rural -0.0389 -0.0455 -0.0616
(0.1014) (0.1006) (0.0988)
Small -0.0801 -0.0773 -0.1047
(0.1781) (0.1787) (0.1721)

61
Large -0.0253 -0.0234 0.0005
(0.0790) (0.0783) (0.0779)
Cumulative Dropout Rate 0.0008 0.0005 -0.0001
(0.0034) (0.0034) (0.0033)
Percent in College Prep Program 0.0003 0.0006 0.0008
(0.0015) (0.0015) (0.0015)
Catholic School -0.0153 -0.0211 0.0237
(0.1701) (0.1706) (0.1761)
Private School -0.2520 -0.2342 -0.1444
(0.3248) (0.3347) (0.3384)

62
Table B4: Descriptive Statistics for Variables in Table 4
Mean SD Min Max
Individual/Family Level Variables (n=5,916)
Ideal and Actual Relationship Script Difference 1.97 0.60 0 5.9
Female 0.51 0.50 0 1
Age 15.41 1.56 11 20
Hispanic 0.15 0.36 0 1
Black 0.21 0.41 0 1
Native American 0.04 0.19 0 1
Asian 0.06 0.23 0 1
Other Race 0.09 0.28 0 1
Multi Race 0.05 0.22 0 1
Home Language Not English 0.09 0.28 0 1
Immigrant 0.07 0.25 0 1
Household Size 4.64 1.62 1 18
Single Parent Household 0.25 0.43 0 1
Other Household Type 0.23 0.42 0 1
Parent Immigrant 0.16 0.37 0 1
Parent Education – HS Grad 0.30 0.46 0 1
Parent Education – Some College 0.30 0.46 0 1
Parent Education -- College 0.24 0.43 0 1
Parent Professional/Managerial Occ 0.34 0.47 0 1
Parent Disabled 0.05 0.21 0 1
Family Welfare Receipt 0.10 0.30 0 1
Log Family Income 3.54 0.85 0 6.91
Low Birth Weight 0.11 0.31 0 1
Mother’s Age at Birth 25.48 5.34 12 53

Neighborhood Level Variables (n=1299)

Neighborhood Disadvantage Scale 0 1 -3.14 3.29
Nhood Heterogeneity Ideal Relationship Script 0 1 -3.22 8.10
Nhood Reliability (weight) 0.66 0.33 0.11 1

School Level Variables (n=143)

Urban 0.31 0.47 0 1
Rural 0.13 0.34 0 1
Small 0.21 0.41 0 1
Large 0.31 0.46 0 1
Cumulative Dropout Rate 7.74 11.18 0 68.52
Percent in College Prep Program 41.44 33.43 0 100
Catholic School 0.03 0.18 0 1
Private School 0.02 0.14 0 1

63
Table B5: Control Variable Coefficients for Models in Table 4
Model 1 Model 2
Individual/Family Level Variables
Female -0.0618* -0.0614*
(0.0142) (0.0142)
Age 0.0221* 0.0216*
(0.0061) (0.0061)
Hispanic 0.0713* 0.0712*
(0.0316) (0.0314)
Black 0.1147* 0.1031*
(0.0231) (0.0252)
Native American -0.0001 -0.0045
(0.0552) (0.0554)
Asian 0.0123 0.0107
(0.0666) (0.0651)
Other Race -0.0068 -0.0099
(0.0371) (0.0371)
Multi Race -0.0252 -0.0188
(0.0630) (0.0638)
Home Language Not English -0.0903 -0.0904
(0.0498) (0.0500)
Immigrant 0.0302 0.0331
(0.0350) (0.0354)
Household Size 0.0008 0.0007
(0.0048) (0.0048)
Single Parent Household 0.0189 0.0200
(0.0293) (0.0294)
Other Household Type 0.0321 0.0331
(0.0226) (0.0226)
Parent Immigrant -0.0052 -0.0040
(0.0335) (0.0334)
Parent Education – HS Grad -0.0738* -0.0720*
(0.0250) (0.0252)
Parent Education – Some College -0.0860* -0.0822*
(0.0258) (0.0259)
Parent Education -- College -0.0882* -0.0836*
(0.0347) (0.0349)
Parent Professional/Managerial Occ -0.0059 -0.0040
(0.0178) (0.0178)
Parent Disabled 0.0096 0.0093
(0.0381) (0.0382)
Family Welfare Receipt 0.0214 0.0208
(0.0305) (0.0305)
Log Family Income -0.0100 -0.0065
(0.0146) (0.0147)
Low Birth Weight -0.0045 -0.0047
(0.0300) (0.0299)
Mother’s Age at Birth -0.0010 -0.0008
(0.0020) (0.0020)
School Level Variables
Urban 0.0001 -0.00405
(0.0194) (0.0194)
Rural -0.0290 -0.03262
(0.0204) (0.0202)
Small -0.0245 -0.02956
(0.0255) (0.0245)

64
Large -0.0298 -0.02323
(0.0201) (0.0211)
Cumulative Dropout Rate 0.0009 0.000608
(0.0007) (0.0008)
Percent in College Prep Program -0.0003 -0.00024
(0.0004) (0.0004)
Catholic School 0.0253 0.029547
(0.0453) (0.0443)
Private School -0.0555 -0.03681
(0.0418) (0.0396)

65
 Appendix C:
Using Sequence Analysis to Measure Differences in Romantic Relationship Scripts

If we think of the ordered elements that constitute a relationship script as a sequence of

events, then an optimal matching algorithm can be applied to measure the difference between

any two relationship scripts. An optimal matching algorithm measures the difference between

two sequences by determining the minimum combination of insertions, deletions, and

substitutions that are required to change one sequence into another. More extensive discussions

of optimal matching and sequence analysis can be found in elsewhere (Abbott 1990, 1995,

Abbott and Hrycak 1990, Sankoff and Kruskal 1983). This discussion is based on those sources

and the summary in Stoval et al (1996).

An example helps to illustrate the idea. The menu of elements that respondents can select

and order in the ideal relationship scripts is:

A: We would go out together in a group

B: I would meet my partner’s parents
C: I would tell other people that we were a couple
D: I would see less of my other friends so I could spend more time with my partner
E: We would go out together alone
F: We would hold hands
G: I would give my partner a present
H: My partner would give me a present
I: I would tell my partner that I loved him or her
J: My partner would tell me that he or she loved me
K: We would think of ourselves as a couple
L: We would talk about contraception or sexually transmitted diseases
M: We would kiss
N: We would touch each other under our clothing
O: We would have sex
P: My partner or I would get pregnant
Q: We would get married

An example relationship script is “EAFMGBHKCLNOIJQP.” Another example is,

“ABEDGMCJILMNOHKPQ.” To change the second script into the first, we can begin by

66
inserting E at the beginning of script two, then substitute F for B, and then substitute M for D.

This will align the first four elements of the scripts so they both start with EAFM. Alternatively,

we could substitute E for A and then A for F, and then insert an F and then an M. Which of these

two sets of operations is optimal (or whether some other is optimal) as well as the cost of the

optimal set of operations depends on the relative costs of insertions, deletions, and substitutions.

The optimal matching algorithm computes the minimum cost of all possible sets of operations

that change one sequence into another given defined costs of insertions, deletions, and

substitutions. This provides a quantitative measure of the difference between two relationship

scripts. I do not discuss the mechanics of the algorithm here. Sankoff and Kruskal (1983) provide

an extensive discussion of such algorithms.

Following Stovel et al. (1996), substitution costs are defined using observed transition

probabilities in the data, such that events frequently following one another in the set of observed

sequences are assigned lower transition costs than those rarely observed following one another.

For example, the event “going out together alone” more frequently follows “going out together

in a group” than it follows “having sex,” so substituting “going out together alone” for “going

out together in a group” has a lower substitution cost than substituting “having sex” for “going

out together in a group.” Here I derive the transition cost matrix from all observed ideal

relationship scripts in the wave one Addhealth data. Substitution costs are defined as the inverse

of the transition probability. To reduce extreme values, substitution costs are then logged using

the natural logarithm. The resulting substitution costs range from 0.70 to 6.97 in the ideal

relationship transition data. The substitution cost matrix is displayed in Table C1. Following

Stovel et al. (1996), insertion and deletion costs are defined as the maximum substitution cost

(6.97). Following Abbott and Hrycak (1990), the difference between two scripts is normalized

67
by dividing by the length of the longer script, so the final difference measure is the average cost

per event in the longer script.

To measure the difference between a particular respondent’s ideal and actual relationship

the optimal matching algorithm is also used. For construction of this measure, substitution costs

are also calculated based on the set of all observed ideal relationship scripts in the wave one

Addhealth data. One difficulty arises in this comparison because, as described in the main text,

the wave one ideal relationships and wave two actual relationships have different sets of events

from which the respondent can choose. The only practical solution is to remove events from the

scripts that are not common to both scripts prior to comparison. A second issue is how to deal

with an actual relationship that has fewer events than a respondent’s ideal relationship. An actual

relationship may end due to decisions made by one’s partner, or the relationship may still be

ongoing at the time of the wave 2 interview. Here I shorten the ideal relationship to the number

of events contained within the actual relationship by removing events at the end of the ideal

relationship script.

68
View publication stats

Table C1: Substitution Cost Matrix for Ideal Relationship Scripts

A B C D E F G H I J K L M N O P Q
A 0.00 1.71 2.32 3.18 1.47 1.85 3.11 3.47 3.79 3.86 2.42 3.69 3.30 4.47 5.70 6.68 5.14
B 2.17 0.00 1.83 3.11 1.87 2.44 2.59 2.86 3.21 3.19 2.33 3.01 3.27 3.93 5.15 6.87 4.29
C 2.47 2.01 0.00 2.38 2.06 1.94 2.70 3.01 3.22 3.21 2.19 3.52 2.90 4.13 5.68 6.92 5.28
D 3.00 2.60 2.68 0.00 1.66 2.57 2.36 2.80 2.85 3.07 2.73 2.58 3.08 2.95 4.40 4.82 3.80
E 2.80 2.73 2.86 3.24 0.00 1.17 2.52 3.05 3.25 3.45 2.37 3.45 2.15 4.03 5.64 6.75 5.26
F 3.08 2.80 2.72 3.97 2.26 0.00 2.12 3.11 3.03 3.49 2.24 4.06 1.13 4.80 6.25 6.97 6.01
G 3.94 3.30 3.99 3.90 3.53 3.36 0.00 0.61 2.93 2.37 3.47 3.15 3.27 3.76 5.24 6.18 5.04
H 3.74 3.13 3.82 3.63 3.41 3.47 1.08 0.00 1.79 2.52 2.82 2.73 2.94 3.38 4.84 5.99 4.58
I 4.34 3.36 3.99 4.06 3.71 3.62 3.06 2.92 0.00 0.69 3.12 2.45 2.92 3.08 4.20 6.02 4.22
J 4.12 3.37 3.80 3.94 3.68 3.48 2.90 3.36 0.94 0.00 2.03 2.37 2.56 3.08 4.50 6.13 4.43
K 3.17 2.78 0.97 3.82 2.99 2.50 3.19 3.49 3.24 3.18 0.00 2.54 2.25 4.27 5.67 6.52 5.18
L 3.80 3.37 3.81 3.55 3.44 3.65 3.47 3.43 3.25 2.85 3.22 0.00 1.98 1.21 1.96 4.79 3.06
M 3.04 2.34 2.74 3.44 2.70 2.85 2.64 2.83 2.73 2.73 2.08 2.44 0.00 2.15 3.89 6.14 3.84
N 4.36 3.85 4.53 4.07 4.81 5.06 3.76 3.98 3.50 3.42 4.31 1.94 3.90 0.00 0.57 3.56 3.08
O 3.53 3.11 3.82 3.04 4.30 4.67 3.16 3.26 3.32 3.07 3.68 3.51 4.41 2.61 0.00 1.29 1.37
P 3.34 3.55 4.10 2.86 4.30 4.70 3.72 3.72 3.93 4.19 4.14 3.16 4.30 2.92 2.01 0.00 0.70
Q 3.73 4.05 4.10 3.00 4.19 4.22 3.82 3.89 3.78 4.19 4.08 2.35 3.44 1.79 1.37 1.53 0.00

69