Field and Practice in American High Schools

1
Field and Practice in American High Schools
James Moody The Ohio State University Popula
tion Research Center University of Texas
Austin April 23, 2004
2

• Introduction
• Cause and Consequence in Social Science
• Put up or shut up empirically testing an
  alternative
• Bourdieu Field
• What is a field?
• From fields to practice, relations, risk
• Analysis
• Data Measures
• How adolescents spend their time
• Activity Profiles as systems of practice
• Global vs. Local perspectives
• Case studies of two High Schools
• Relational position through blockmodels
• Correspondence of Practice Position
• Practice, Position Youth Risk Behavior
• Tentative conclusions
• Future (re)directions

3
Cause and Consequence in Social Science

• Standard view
• Science seeks to identify causes of various
  sorts, with an eye toward manipulation.
• The standard requirements for causal inference
  are (1) isolation, (2) association, and (3)
  direction (Bollen, 1989).
• The most common methods are variations of the
  general linear model, either singly or in systems
  of linked equations.
• Critiques are common
• Within the standard view there are serious
  debates about drawing causal conclusions from
  non-experimental data Lieberson (1989) Making it
  Count, Perl Causality Sobel (1989, 1997, 1996).
  Concerns basically amount to repeated attempts to
  better specify either isolation, association, or
  direction peppered with concerns for endogeneity
  and selection.
• Outside the standard view, authors argue that
  treating cause as a system of intersecting forces
  misrepresents the fundamental actively embedded
  nature of social life. Abbott has probably been
  the most visible critic on this front ().

4
(No Transcript)
5
What is a Field?
Field serves as some sort of representation for
those overarching social regularities that may
also be visualized as quasi-organisms, systems
or structures J. L. Martin AJS
2003. Bourdieus notion of field particularly
as discussed in Language and Symbolic Power can
be thought of as a space of organized striving
- a social topology composed of various
dimensions of capital that are relevant in a
field. This implies two related aspects of a
field (1) the dimensions of capital used to
construct the field and (2) the practical guides
that generate action in a given activity domain.
Examples of fields range from abstract notions
of status spaces to concrete examples such as the
French academic system.
6
What is a Field?

• Martin identifies 3 basic dimensions of a
  (socoiological) definition of field
• A topology ? a space of/for action, but not
  simply the distributions of attributes (I.e. not
  the same as Blau space)
• A field of forces ? thought of as vectors in the
  space that act on participants in the field. The
  analogy here is of magnetic or gravitational
  fields.
• A battlefield ? a place of active, organized
  contestation for the resources (usually some sort
  of status) at play in the field.
• We can capture elements of each of these by
  identifying the relation between position in a
  social network and the practices that people
  engage in to signal their position in the field.

7
Why Look For Fields?

• What are some advantages to conceptualizing
  social spaces as fields, rather than focusing on
  more normal-science explanations for behavior?
• Simplification. Field concepts provide a way to
  undstand systemic activity patterns in large
  collectives at a distance. Just as iron
  filings respond to magnetic fields producing a
  clear pattern relative to each other, peoples
  activities (theoretically) admit to regularities
  that cannot be explained by standard
  variable-centered approaches.
• Embedded Action ontology. Field concepts allow us
  to move beyond questions about the essence of
  variables as active agents to notions of people
  actively navigating community structures.
• Relational Focus. Field concepts effectively
  move the unit of analysis from an individual
  actor to sets of actors who are positioned
  relative to each other through social relations
  and behaviors.
• There are also costs. Field concepts are
  notoriously difficult to specify, causal notions
  do not follow social science standards for
  explaining variance (but in so doing allows an
  analysis of constants), and some have argued that
  they slip uncomfortably toward tautology.

8
Fields in High School

• Weve known since at least Elmtown's Youth and
  certainly since The Adolescent Society that
  schools are significant sites for status
  struggles.
• Popularity, the Leading Crowd Thugs, and
  so forth all signal positions that carry status
  implications.
• Ethnographic work in schools suggests that youth
  are actively engaged in exploiting their
  behaviors and relations to position themselves
  within this game.
• Behaviors, dress, etc. signal a particular
  position
• Adolescents actively manage their social
  relations for status concerns.
• Relations themselves are of key interest to
  adolescents
• The logic of practice in such fields suggest that
  position in the field should correspond to a
  relational structure and a distinct pattern of
  how kids spend their time.

9
From Field to Practice

• Bourdieu seems to suggest that position in a
  field results in a distinct set of practices
  that, combined, define relative positions in a
  given field. These practices are keyed to
  everyday experience, and importantly take place
  in the real-life constrained time available to
  participants.
• I attempt to operationalize positions of practice
  by identifying characteristic sets of everyday,
  time-consuming, activities.
• I use (endogenously determined) sets of
  activities for 2 reasons
• Descriptively, sets might provide a simpler way
  to characterize ideal types
• Behaviorally, I suspect that adolescents collapse
  others into idealized categories (though they
  probably have fuzzy boundaries).
• The alternative is to measure distance without
  creating sets.
• (Note this might well be contested. Part of the
  reason I shy away from habitus directly is the
  link between habitus and dispositions, which (a)
  are even more difficult to measure than practices
  and (b) Im not convinced that actors necessarily
  are conscious of, and thus able to answer
  meaningfully about, such dispositions)

10
From Field to Relations
The discussion of the relational foundation for
social fields suggests a clear network
linkage. Most of his discussion of the
relational foundation of fields rests on ideas
that seem very similar to roles a relatively
small number of positions defined by
characteristic involvement in patterned social
relations. Notions of field (in Bourdieu or
otherwise) have used structural equivalence to
operationalize fields.
11
Data Methods

• Data
• I use the National Longitudinal Survey of
  Adolescent Health (Add Health). This is a
  nationally representative survey of adolescents
  (7th through 12 grade). My sample consists of
  students who filled out both the in-school and
  wave-1 in-home survey. I use the two largest
  saturated schools as case studies.
• Methods
• Practice Positions are defined using cluster
  analysis across 17 time-consuming, everyday
  activities that youth report being involved in.
• Global versions situate each student relative to
  all others in the sample
• Local versions identify clusters by comparing
  people only to those in their own schools
• Relational Positions are defined using block
  models based on regular equivalence
• The correspondence of position and practice is
  determined through relational (dyad) models.
• The effect of position and practice on
  risk-related behaviors uses multivariate
  regression techniques.

12
Adolescent Activities
13
Adolescent Activities By Sex
14
Adolescent Activities By Grade in School
To generate activity sets, I cluster over these
17 activity classes. To avoid simply
developing markers for gender and age, I
standardize all variables within grade and sex
15
Global Cluster Solution
3007
5007
2000
7950
2365
2943
578
901
2413
512
6222
1998
3809
1811
Cluster tree based on a Wards minimum variance
clustering on 17 variables, standardized by age
and sex.
16
Global Cluster Solution
Cluster 1 Low Activity, Church Going, Drivers
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N3007 (21)
17
Global Cluster Solution
Cluster 2 Low Activity, Workers
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N2000 (14)
18
Global Cluster Solution
Cluster 3 High Activity, Clubs
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N2943 (21)
19
Global Cluster Solution
Cluster 4 TV Video Games
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N901 (6)
20
Global Cluster Solution
Cluster 5 Non-School Active
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N512 (3.5)
21
Global Cluster Solution
Cluster 6 Church Going, Non-workers
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N1998 (14)
22
Global Cluster Solution
Cluster 7 Inactive Uninvolved
1.5
1
0.5
0
TV
Drive
Chores
Church
Cycling
Hobbies
Exercise
Hang Out
V.Games
Arts_club
sport_club
HrsWork
Active Sport
Yth. Grps
slife_club
PaidWork
Acad_club
-0.5
-1
-1.5
N1811 (13)
23
Local Practices Case Studies of Sunshine
Jefferson high schools
Center of the school district. Streets and
boundaries are educated guesses. Up is not
necessarily North.
24
Local Practices Case Studies of Sunshine
Jefferson high schools
Center of the school district. Streets and
boundaries are educated guesses. Up is not
necessarily North.
25
Local Cluster Solution Jefferson High School
Cluster 1 Non- Active, Workers
129
324
Cluster 2 Non- School Active
195
488
106
Cluster 3 Non-working, Non Religions
164
58
Cluster 4 Video Gamers
98
Cluster 5 Church Going, School Active
108
10
Cluster tree based on a Wards minimum variance
clustering on 17 variables, standardized by age
and sex.
26
Local Cluster Solution Jefferson High School
Cluster 1 Non-active, Workers
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N129 (22)
27
Local Cluster Solution Jefferson High School
Cluster 2 Non-School, active
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N195 (33)
28
Local Cluster Solution Jefferson High School
Cluster 3 Non-workers, non-religious
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N106 (18)
29
Local Cluster Solution Jefferson High School
Cluster 4 Video Gamers
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N58 (10)
30
Local Cluster Solution Jefferson High School
Cluster 5 Church Going, School Active
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N108 (18)
31
Local Cluster Solution Sunshine High School
Cluster 1 Active Non-arts
235
Cluster 2 Non- School, Active
718
277
483
Cluster 3 Inactive Uninvolved
1064
206
346
Cluster 4 Workers
Cluster 5 School Active, Artists
99
Cluster tree based on a Wards minimum variance
clustering on 17 variables, standardized by age
and sex.
32
Local Cluster Solution Sunshine High School
Cluster 1 Active, Non Arts
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N235 (20)
33
Local Cluster Solution Sunshine High School
Cluster 2 Non-School Active
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N272 (23)
34
Local Cluster Solution Sunshine High School
Cluster 3 Inactive Uninvolved
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N206 (18)
35
Local Cluster Solution Sunshine High School
Cluster 4 Workers
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N346 (30)
36
Local Cluster Solution Sunshine High School
Cluster 5 School active, Artists
1.5
1
0.5
0
TV
arts
drive
Cycle
Sport
Work
sports
Chores
Church
std_life
Friends
Hobbies
hrswork
Exercise
videogames
Youthgrps
acad_club
-0.5
-1
-1.5
N91 (8)
37
Comparing Local to Global
The overlap between local and global assignments
is statistically significant (adjusted rand of
about 15), but the results suggest a fair amount
of room for local variation on the global theme.
This suggests that, while there are some
similarities, social relations might admit to
patterning via a local embeddedness or a global
identity.
38
Network Structure

• Bourdieus discussion of relational fields
  suggests that position in the fields status
  structure will correspond to the development of
  unique habitus, and thus distinct practices.
  This leaves open (at least) two possible ways to
  operationalize position in the field.
• Position as regular equivalence.
• If position relates to a generalized notion of
  status, then two actors with similar patterns of
  relations in the network should have similar
  behavior profiles. The mechanism here suggests
  aligning status interests and mimicry between
  those in analogous positions in the field.
• Position as direct connection between actors.
• If one learns the relevant action rules through
  direct interaction with others in the field, then
  those people who are directly connected should
  have the most similar behavior profiles. The
  mechanism here is largely one of communication,
  peer influence selection.

39
Network Structure Regular Equivalence through
triad distributions
(0)
(1)
(2)
(3)
(4)
(5)
(6)
003
012
102
111D
201
210
300
021D
111U
120D
Intransitive
Transitive
021U
030T
120U
Mixed
021C
030C
120C
40
Network Structure Regular Equivalence through
triad distributions
Type Number of
triads ---------------------------------------
1 - 003 21 ----------------------
----------------- 2 - 012 26
3 - 102 11 4 - 021D
1 5 - 021U 5 6 -
021C 3 7 - 111D
2 8 - 111U 5 9 - 030T
3 10 - 030C 1
11 - 201 1 12 - 120D
1 13 - 120U 1 14 -
120C 1 15 - 210
1 16 - 300
1 --------------------------------------- Sum (2
- 16) 63
41
Network Structure Regular Equivalence through
triad distributions
PRC300,102, 003, 120D, 120U, 030T, 021D, 021U
Ranked Cluster
1
0
0
0
0
1
1
0
0
0
1
1
0
0
0
1
1
1
1
0
1
1
1
1
0
And many more...
42
Network Structure Regular Equivalence through
triad distributions
43
Block Model School Network Structure
Jefferson High School
Sunshine High School
School provides a good boundary for social
relations
School does not provide a good boundary for
social relations
44
Block Model Jefferson High School
64
1 Outsiders
276
Semi- periphery
212
2 Insiders
544
3 Outsiders
147
Periphery
268
4 Insiders
121
5 Aloof
80
6 Most Popular
87
Core
288
7 2nd String
121
Wards minimum variance clustering on the 36
triad types, plus number of ties sent
out-of-school
45
Block Model Sunshine High School
253
Receiving Periphery
432
179
903
Sending Periphery
421
471
50
Semi- Periphery
255
742
Lieutenants
332
818
487
102
155
53
76
Core
Wards minimum variance clustering on the 36
triad types, plus number of ties sent
out-of-school
46
Block Models
Jefferson High School
Sunshine High School
4
34
43
32
52
33
Image networks. Width of tie is proportional to
the ratio of cell density to mean cell density.
47
 
Block Model Characteristics
Jefferson High School
Sunshine High School
 
48
Block Model Sunshine High School
Jefferson High School
Sunshine High School
49
Correspondence of Field Practice
Jefferson High School

• Being in the same block significantly increases
  the likelihood of being the same behavioral
  cluster
• Locally defined OR 1.13
• Globally defined OR 1.12
• The effect is differential across blocks
• Being adjacent in the network has a consistent
  positive effect
• Local OR 1.21
• Global OR 1.35
• SES Similarity increases the odds of being in the
  same behavior profile

Block Local Global
1 Semi P Outsiders 1.03 1.16
2 Semi P Insiders 0.89 0.83
3 Periphery Outsiders 1.76 2.08
4 Periphery Insiders 1.03 1.09
5 Aloof Core 1.15 0.97
6 Popular Core 1.01 0.79
7 2nd string Core 1.08 1.04
Coefficients based on a dyad-level logistic
regression model. Models control for grade and
gender.
50
Correspondence of Field Practice
Sunshine High School

• Being in the same block barely increases the
  likelihood of being the same behavioral cluster
• Locally defined OR 1.03
• Globally define OR 1.02
• The effect is differential across blocks
• Being adjacent in the network has a weaker, but
  still positive effect
• Local 1.13
• Global 1.08
• SES Similarity increases the odds of being in the
  same behavior profile

Block Local Global
1 Receiving Periphery 1.25 1.27
2 Sending Periphery 0.99 1.01
3 Semi Periphery 0.93 0.89
4 Lieutenants 0.93 0.88
5 Popular Core 0.87 1.08
Coefficients based on a dyad-level logistic
regression model. Models control for grade, race
and gender.
51
Correspondence of Field Practice

• Jefferson High School
• Periphery Outsiders are most likely to be members
  of the Video Gamers local cluster and the Low
  Activity, Working global cluster.
• Aloof core members are most likely to be members
  of the Non-School, Active local cluster.
• Sunshine High School
• Receiving Periphery members are most likely to be
  in the Workers local cluster and the Low
  Activity, Church Going global cluster. My sense
  is this turns on Drivers
• But the relation is far from direct, and simple
  adjacency is a stronger predictor.

52
Field, Practice Adolescent Risk Behavior
Delinquency
Scale ranges from never to 6 or more times
and is highly skewed. I dichotomize to identify
the 25 who are most delinquent
53
Field, Practice Adolescent Risk Behavior
Sexually Active
Proportion who report having had sex
Grade in school
54
Field, Practice Adolescent Risk Behavior
School Attachment
Mean of 3 items - Happy to be at school -
Feel a part of the school - Feel close to
others at school
55
Field, Practice Adolescent Risk
Behavior Jefferson High School Delinquency

• Core blocks in Jefferson are much more likely to
  be delinquent than either of the insider
  periphery blocks.
• Periphery Insiders are 0.32 times as likely to be
  delinquent
• Semi-Periphery Insiders are 0.53 times as likely
  to be delinquent

Models control for SES, Sex, grade in school and
family structure, values are odds rations for
being delinquent.
56
Field, Practice Adolescent Risk
Behavior Jefferson High School Delinquency

• Based on the Local activity profile, students in
  the Non-active working group are significantly
  less likely (OR 0.309) to be delinquent.
• Based on the Global activity profile, students in
  the Non-School Active group are significantly
  more likely to be delinquent.
• Combining block membership and behavior profiles
  suggest that the three sources are largely
  independent, with each making significant
  contributions to the likelihood of delinquency.
• The substantive effect of the network position
  variables seems highest
• Followed by the effect of local behavior profile
• The global behavior effects are the least
  powerful.

Models control for SES, Sex, grade in school and
family structure
57
Field, Practice Adolescent Risk
Behavior Jefferson High School Sexual Activity
Core block members are more likely to be sexually
active than outsiders, with members of block 1
being particularly unlikely to have had sex.
Models control for SES, Sex, grade in school and
family structure, values are odds ratios for
being sexually active.
58
Field, Practice Adolescent Risk
Behavior Jefferson High School Sexual Activity

• Based on the local activity profile, students in
  the Non-active working group are significantly
  more likely (OR 2.13) to report having had sex.
• Note that this correlation is opposite the
  general positive relation between sexual activity
  and delinquency in Add Health as a whole.
• The global activity profile neatly sorts groups
  according to how likely they are to have sex
• Low activity workers (Or3.56) and Inactive
  Uninvolved (1.75) are much more likely to report
  being sexually active.
• Church going non-workers are much less likely
  to report being sexually active (OR 0.19).
• Combining block membership and behavior profiles
  suggest that the three effects are largely
  independent, each making significant
  contributions to sexual activity, though
• The effect of global activity profile seems the
  strongest
• Followed by block position
• Then by local activity level

Models control for SES, Sex, grade in school and
family structure
59
Field, Practice Adolescent Risk
Behavior Jefferson High School School Attachment
Periphery Insiders feel the least attached to
school and core groups 6 7 are most attached,
net of other factors.
School Attachment
4
3.8
3.6
Mean School Attachment
3.4
3.2
Semi- Periphery Outside
Semi- Periphery Inside
Periphery Outside
Periphery Inside
Core 2nd string
Core Popular
Core Aloof
3
2.8
1
2
3
4
5
6
7
Block
Models control for SES, Sex, grade in school and
family structure
60
Field, Practice Adolescent Risk
Behavior Jefferson High School School Attachment
School Attachment
4.1
4
3.9
3.8
3.7
3.6
3.5
3.4
3.3
3.2
3.1
3
1 Non active Workers
2 Non-School Active
3 Non-working Non Religious
4 Video Gamers
5 Church Going School Active

• Adding block position removes the effect of being
  a gamer
• Global position shows that
• Group 1 (Low activity church going) are more
  attached
• Group 5 (Non-school active) are less attached
• The network position effects and local cluster
  effects are the strongest

Models control for SES, Sex, grade in school and
family structure. Cross color comparisons are
statistically significant.
61
Field, Practice Adolescent Risk
Behavior Sunshine High School Delinquency
Core Periphery members are both delinquent,
though I suspect they are involved in different
activities
Models control for SES, Sex, grade in school and
family structure, values are odds ratios for
being delinquent.
62
Field, Practice Adolescent Risk
Behavior Sunshine High School Delinquency
Local activity effects are weak, but those in the
Non-School Active cluster (2) are slightly more
likely to be delinquent (OR 1.32). The global
activity profile suggests a somewhat more
differentiating effect Local
behavior profile effects dissolve after adding
global effects and network position, though they
remain largely intact, with periphery actors
having the strongest network effect.
Models control for SES, Sex, grade in school and
family structure, values are odds ratios for
being delinquent
63
Field, Practice Adolescent Risk
Behavior Sunshine High School Sexual Activity
Core members are the least likely to be sexually
active. Those on the periphery are most likely.
Models control for SES, Sex, grade in school and
family structure, values are odds ratios for
being sexually active.
64
Field, Practice Adolescent Risk
Behavior Sunshine High School Sexual Activity
Students in the Worker cluster are more likely
to report being sexually active than any of the
other clusters (OR 1.45 respectively). The
global activity profile suggests that those in
Non-School Activities are much more likely
(OR3.28) to be sexually active, while
high-activity club students and those who are
Inactive uninvolved are much less likely to
be sexually active (OR 0.46 0.68
respectively). Global profile behavior account
for Receiving Periphery position
effects. Combined, the global effects seem to
supercede local effects, but network position
remains strong. Core members are somewhat less
likely to be sexually active sending periphery
members are more likely to be sexually active,
net of behavior activities.
Models control for SES, Sex, grade in school and
family structure
65
Field, Practice Adolescent Risk
Behavior Sunshine High School School Attachment
School Attachment
4.3
4.2
4.1
4
3.9
Mean Attachment
3.8
3.7
3.6
3 Semi-Periphery
5 Popular Core
Receive Periph.
4 Lieutenants
Send Periph.
3.5
3.4
3.3
Cross-color comparisons are statistically
significant
Models control for SES, Sex, grade in school and
family structure
66
Field, Practice Adolescent Risk
Behavior Sunshine High School School Attachment
Local activity profiles suggest that those who
are Inactive Uninvolved Workers are
significantly less attached to school. This
effect of local position are independent of the
network position effects and largely
additive. Global activity profiles suggest that
those who are in the high activity club profile
and TV / Video Gamers profile are more likely
to be attached to school, while those who are in
Non-school active profile are least likely to
be attached to school. All of these effects are
independent of each other and remain strong when
entered simultaneously into the model.
67
Field, Practice Adolescent Risk
Behavior Tentative Conclusions Local or Global
Practice Profiles?

• For Jefferson, where the school seems to capture
  student social life, local behavior profiles fit
  much better than in Sunshine, where the school
  does not effectively capture student social
  relations.
• This is more true for delinquency school
  attachment than for sexual activity, which seems
  more affected by position in the global profile.
• Looking at the global profiles, three stand out
  as having particularly interesting effects
  High-Activity Clubs, Non-School but Active, and
  Inactive Uninvolved. Interestingly, the
  effects differ across the two schools.

68
Field, Practice Adolescent Risk
Behavior Tentative Conclusions Local or Global
Practice Profiles?
In Jefferson Low Delinquency In Sunshine Low
Delinquency Low Sexual Activity Most Attached
to school
N2943 (21)
69
Field, Practice Adolescent Risk
Behavior Tentative Conclusions Local or Global
Practice Profiles?
In Jefferson Most likely to be
delinquent Least attached to school In
Sunshine Most delinquent Most sexually
active Largely Unattached
N512 (3.5)
70
Field, Practice Adolescent Risk
Behavior Tentative Conclusions Local or Global
Practice Profiles?
In Jefferson Most likely to be sexually
active Least likely to be delinquent In
Sunshine Unlikely to be sexually active Least
likely to be delinquent
N1811 (13)
71
Field, Practice Adolescent Risk
Behavior Tentative Conclusions How do networks
matter?
In both schools, the networks have a basic status
hierarchy structure with a clear distinction
between core and periphery, which seems important
for risk outcomes.
Core Periphery
Jefferson 2/3 are attached to school Aloof core less so Most sexually active Core blocks are delinquent Inside-oriented are not attached Inside-oriented are least delinquent
Sunshine Most attached to school Least sexually active Delinquent (weak effect) Least attached to school Most sexually active More delinquent
72
Field, Practice Adolescent Risk
Behavior Tentative Conclusions Correspondence
between relations and behavior

• The weak between relational position and behavior
  profiles is there, but its weak.
• That both relations and behaviors affect risk
  outcomes suggests that I might be missing a key
  aspect of the behavior profile. Risk behaviors
  should probably be considered as part of the
  profile.
• The strength of the distinction between core
  periphery AND the that adjacency is a better
  predictor of behavior suggests that we should
  conceive of network position in a way that
  accounts for connectivity.
• Cohesive Blocking (Moody White) seems a good
  possibility
• -Should reconsider the meaning of a friendship
  nomination
• - Distinguish ties within and between sex and
  grade
• - Account for ties that include activity values

73
Field, Practice Adolescent Risk Behavior Future
(re)directions

• The analysis thus far suggests a couple of
  potential avenues that might be more profitable
• The current model tries to split the difference
  between a relational approach and a more standard
  social science model. Instead, it might make
  sense to use a completely relational model.
• Suggests treating the entire exercise as modeling
  distances between actors in behavior space and
  relational space. This drops the problems
  associated with cluster analysis, but makes it
  harder to talk directly to policy.
• Expand the definition of behaviors to include
  dispositions. This raises more measurement
  concerns, but fits with Bourdieus general frame.
• Use the physical distance as a marker for
  unmeasured field factors relate the geographical
  distance directly to the social distance.
• Incorporate a model of trajectories. The current
  model looks at a cross-section of both relations
  and behaviors, but adolescents are surely moving
  through both of these spaces at different clips.
  Is there a satisfactory way to model the
  correspondence in trajectories through these two
  spaces?

74
Field, Practice Adolescent Risk
Behavior Tentative Conclusions