Survey Methods for Network Data

In: The SAGE Handbook of Social Network Analysis

By: Peter V. Marsden

Edited by: John Scott & Peter J. Carrington
Pub. Date: 2016
Access Date: October 20, 2018
Publishing Company: SAGE Publications Ltd
City: London
Print ISBN: 9781847873958
Online ISBN: 9781446294413
DOI: http://dx.doi.org/10.4135/9781446294413
Print pages: 370-388
© 2014 SAGE Publications Ltd All Rights Reserved.
This PDF has been generated from SAGE Research Methods. Please note that the
pagination of the online version will vary from the pagination of the print book.
SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Survey Methods for Network Data

Peter V.Marsden

Introduction

Surveys and questionnaires are widely used to assemble data on connections among persons
or other social actors. Such methods have a long heritage: researchers used them to study
interpersonal relationships among children as early as the 1920s (Freeman, 2004). In the
twenty-first century, archival measures of linkages and transactions based on administrative
records or computer-mediated communication systems have become much more abundant and
accessible. Surveys nonetheless remain a vital source of network data for the many situations
in which such records do not exist or do not include information about the relationships of
interest, or in which direct observation, diaries, and other methods of collecting network data
are impractical.

This chapter reviews basic issues in conducting network surveys, beginning with the common
forms that survey network data take, approaches to defining the boundaries of the study
population, and methods of selecting subjects/respondents. It then discusses instruments for
data collection in both whole-network and egocentric designs, illustrating different approaches
with empirical studies. Next, it reviews some cognitive considerations to keep in mind when
developing and administering network surveys. The chapter closes by reviewing some research
into the quality of network data obtained using survey methods and by highlighting special
issues of human subject protection that can arise in the course of conducting network surveys.

Templates for Network Data

Two principal types of network surveys focus on different, though related, objects of
measurement: “whole” and “egocentric” networks. Whole-network studies seek to measure the
structure of some bounded social group by collecting data on one or more types of
relationships that link the units or actors within the group. Egocentric network studies have the
more limited objective of describing local social environments by measuring the relationships in
the vicinity of one or more focal units or actors. Whole-network studies subsume egocentric

ones, in a sense: whole-network data include an egocentric network for each actor.1

Most often, surveys to measure whole networks collect one-mode network data on relationships
among elements of a single set of units or actors – such as friendships among students in a
school, or collaborative ties linking employees in an organization. Sometimes, however, they
measure two-mode networks based on relationships among elements in two distinct sets;

Page 2 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

examples include school networks consisting of student memberships in groups such as

extracurricular clubs or athletic teams, and organizational networks defined by employee
assignments to committees or project teams.

Many design variations arise within these broad study templates. The minimal design measures
only a single type of relationship, such as friendship or advising, on a single occasion.
Extensions include designs that measure relationships among a set of actors on multiple
occasions and designs that measure multiple types of ties (e.g., friendship, advising, and
collaboration) at a single time point. Most studies supplement network data with information on
attributes of units/actors, attributes of dyadic ties, or both. Studies that measure networks for
two or more groups may also include group-level attributes.

Defining and Sampling the Study Population

Defining the target population to be described and constructing a sampling frame that
enumerates (or otherwise provides access to) it are early decisions in conducting social
surveys. For network studies, these decisions about the population of elements and linkages to
be measured are known as the “boundary specification problem” (Laumann et al., 1989).

Whole-Network Studies

For some whole-network studies, formal or positional criteria offer relatively self-evident
definitions of inclusion within the group of interest. Examples include employment by a
physician practice (Keating et al., 2007), assignment to a school classroom (Hansell, 1984),
and residence in a district (Kirke, 1996).

Delineating boundaries can be more challenging for groups that are not formally defined, such
as “regulars” at a beach (Freeman and Webster, 1994), policy domains (Laumann and Knoke,
1987), or social services delivery systems (Doriean and Woodard, 1992). Positional criteria can
help to identify some actors in such populations. Observing (or referring to records
documenting) participation in some set of relevant events also can be useful. For example,
Laumann and Knoke's (1987) study identified some organizational participants in the U.S.
health and energy policy domains based on appearances before relevant congressional
committees, filing of amicus curiae briefs with appellate courts, and registration of lobbyists,
among other criteria. Asking knowledgeable informants to nominate participants on a
reputational basis (prior to data collection) can supplement positional and event-based
approaches to boundary specification.

Social relationships themselves can also indicate inclusion in a population, while their absence

Page 3 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

can signal group boundaries; Laumann et al. (1989) term this a “relational” strategy for
boundary specification. For example, a Colorado Springs study of persons engaged in high-risk
behaviors (Klovdahl et al., 1994) enrolled partners of prostitutes and injecting drug users in its
study population. Studies of service delivery systems could begin with some core agencies,
later adding others to which they refer clients (Doreian and Woodard, 1992). Relational criteria
are often used to add actors to the study population during the course of fieldwork, on the
basis of ties to or nominations made by early respondents. Positional, event-based, reputational
and relational criteria can be used together to identify populations for network studies; a study
might begin with a list of actors included on a positional basis, and then supplement it using
event-based or reputational criteria before fieldwork begins.

The process of boundary specification often results in a complete listing or roster of the study
population. This can be an important aid to whole-network data collection.

Egocentric Network Studies

When egocentric network studies are conducted as part of representative sample surveys (e.g.,
Marsden, 1987), boundary specification follows the definition of the target population for those
surveys. A second boundary-determination problem for egocentric studies has to do with
delimiting the set of “alter” actors within any given respondent's egocentric network. In practice,
“name generators” used in eliciting egocentric network data (see below) often serve this
purpose.

Network Sampling

In contrast to many social science surveys, network surveys only sometimes draw samples.
Whole-network studies very often attempt to collect data on relationships among all actors in a
population, for example, because they seek measures of relational properties (Lazarsfeld and
Menzel, 1980) for all elements. When data collection uses survey methods, this is realistic only
for populations of small to moderate size. Egocentric network data are often obtained within
sample surveys, relying on their sampling methods to select a representative set of focal actors.

Some structural properties of whole networks may be estimated by sampling from networks.
Samples may be drawn by selecting units or actors, or by selecting relationships. Sampling
actors is usually a practical choice for surveys, since they assemble data by contacting actors.
Frank (1981) describes several network sampling schemes. One approach draws a probability
sample of actors (e.g., by using a simple-random or Bernoulli scheme), and then observes only
those relationships within the subset of sampled actors. Another draws a probability sample of

actors and then observes all relationships incident to those actors.2 Networks may also be

Page 4 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

sampled by link-tracing methods that begin with a probability sample of actors, elicit their
contacts, and subsequently sample the contacts (e.g., Liebow et al., 1995). Such link-tracing
may then be repeated one or more times to yield sampled “random walks” containing three or

more linked actors.3

Different inferences about network properties are available from different sampling designs, so
one must consider carefully what properties of a network are to be estimated when designing a
network sampling scheme. The literature on network sampling (Frank, this volume) should be
consulted at this stage.

Instruments for Measuring Networks

This section introduces approaches commonly used in standardized questionnaires and

interviews to obtain data on social relationships, beginning with methods for measuring whole
networks. It then turns to techniques for measuring egocentric networks, concentrating on
“name generator” instruments that yield the most extensive network data. Last, the section
introduces some shorter instruments: global survey items and multiple-item instruments that
measure one or more specific egocentric network properties but do not elicit reports about
specific actor-to-actor relationships.

Instruments for Whole-Network Data

Measuring whole networks requires information sufficient to assign a value to the relationship
aij between each (ordered) pair of actors i and j, i ≠ j, within a network. We concentrate here on
instruments for one-mode networks.

The Sociometric Test

Surveys to assess whole networks usually administer some variant of the sociometric test
developed by Moreno (1953). The basic technique asks each person i within a network to
identify the “alters” (j) with whom he or she has – or would like to have – a given type of
relationship, yielding a value of aij based on i's “choice” or “non-choice” of j. Originally
developed to measure preferences to associate with (or to avoid) others, the sociometric test
generalizes to measuring “actually existing” relationships (e.g., communication, friendship,
support). Likewise, it can be readily adapted to measuring two-mode networks by asking
respondents to report group memberships or other affiliations of interest. Sociometric items
have been administered in most survey modes, especially in-person interviews and self-
administered questionnaires. Many contemporary applications use computer-assisted modes,
which can simplify presentation of questions and data processing.

Page 5 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Studies use varying criteria – guided by the substantive questions they pose – to elicit
sociometric choices. Some examples of such questions appear in Table 25.1. Keating et al.
(2007) surveyed primary care physicians about their influential conversations about women's
health issues. Singleton and Asher (1977) questioned third-grade students about who they like
to play with and (separately) who they like to work with. Espelage et al. (2007) asked seventh-
graders to name others whom they “hang out with most often” at school. Laumann and Knoke
(1987) asked informants representing organizations in the U.S. energy policy domain to
indicate other organizations in the domain with which their organization “regularly and routinely
discusses national energy policy matters.” Many studies elicit two or more types of contact;
Brass (1985), for example, asked employees at a newspaper publishing company about
workflow ties, work-related communication, and close friendships.

Table 25.1 Examples of whole-network measurement tasks

A.Single-criterion recognition question (Keating et al., 2007)

Please circle the number of conversations that you have had with each of the following
[clinic name] primary care physicians in the last six months that have influenced your
thinking about women's health issues.
[followed by alphabetized list of physicians and response categories ‘0’, ‘1–3’ and ‘≥4’]
B.Multiple-criterion recognition questions (Singleton and Asher, 1977)
How much do you like to play with this person at school?
How much do you like to work with this person at school?
[presented within roster listing students in a class alphabetically; responses were
numbers 1–5 accompanied by faces ranging from frowning to smiling]
C.Free-recall question (Coleman, 1961)
What fellows here in school do you go around with most often? (Give both first and last
names)
[from boys’ version of questionnaire; girls received a questionnaire with slightly different
wording]
D.Cognitive social structure task (Casciaro et al., 1999)
By putting an X in the cells of the following matrix, please indicate whether you think the
people listed in each row consider the people listed in each column as personal friends.
For example, if you think that Ms. J (row 9) considers Mr. N (column N) as a friend, place
an ‘X’ in the corresponding cell ‘9N.’
[followed by square matrix listing persons, with solid shading in diagonal (self-relation)
cells]
E.Social-cognitive mapping task [free recall] (Cairns et al., 1985)

Page 6 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Now tell me about your class: Are there some people who hang around together a lot?
Who are they?
Are there some people who don't hang around with a particular group? Who are they?

Many surveys that administer sociometric items supply a roster of possible alters in the network
for respondents to consult when naming associates. Examples of studies that use such
recognition techniques include Hansell's (1984) study of fifth- and sixth-grade classrooms,
Lazega's (2001) study of attorneys in a law firm, and Provan et al.'s (2009) study of agencies
serving the severely mentally ill in an Arizona county. Other versions ask respondents to freely
recall their ties from memory, as in Coleman's (1961) study of networks within high schools,
Brass's (1985) study of a newspaper publishing company, or Burt's (2004) study of supply
chain managers.

Rosters simplify the reporting task by reminding respondents of the eligible alters within the
network. Using rosters limits measurement error due to the forgetting of associates
documented by Brewer (2000). Sudman's (1985) experiments demonstrated that recognition
methods yield larger networks (see also Hlebec and Ferligoj, 2002). Reviewing and considering
all names on a large roster can be a cumbersome, tedious task for respondents, however. Care
with names of alters is warranted, using either approach: recall methods must ensure that
citations of alters known by different names (e.g., nicknames, titles, spelling variations) are
correctly matched, while rosters used to aid recognition must use the names by which persons
are actually known.

Some early guidelines for sociometric measurement recommended that respondents be

permitted to make an unlimited number of choices (Lindzey and Borgatta, 1954), but others
suggested a limit of three or four citations (Northway, 1952). Network surveys using such limits
include Coleman et al.'s (1966) study of physicians and Laumann and Pappi's (1976) study of
community leaders. Limits have practical advantages in survey administration: they simplify and
specify a sociometric task for respondents, thereby reducing burden. Measurement error
considerations also arise, however. Imposing limits can induce both false negatives (if a
respondent's actual number of associates exceeds the limit) and false positives (if respondents
are encouraged to cite additional alters in order to reach the limit). Bias is thereby produced in
many basic network structure statistics including the degree distribution(s), network summaries
such as the dyad and triad censuses (Holland and Leinhardt, 1973), and others.

Sociometric items measure relationships using diverse sets of response categories and formats.
Binary measurement is very common: respondents may indicate those alters with whom they
have a given type of contact by listing them, marking them on a roster or checklist, or making a

Page 7 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

separate “yes/no” response about each one. The latter “forced-choice” format is more time-
consuming but may encourage deeper processing and more thoughtful answers (Smyth et al.,
2006). Many studies request ordinal assessments: Keating et al.'s (2007) study of physicians
used a three-category frequency scale (0, 1–3, or 4+ conversations); Fernandez (1991)
measured respect relationships in a public finance agency using a five-point scale (“very little”
to “very much”); and Johnson and Orbach (2002) asked political actors to rate their interactions
with one another on an 11-point scale. In some studies of elementary school students or
adolescents (Singleton and Asher, 1977; Hansell, 1984), icons supplement ordinal response
categories: a face with a broad smile indicates that the respondent likes an alter “a lot,” for
example.

Cognitive Social Structure Task

Typical sociometric items ask respondents to report only on relationships in which they are
directly involved. A cognitive social structure (CSS) design (Krackhardt, 1987) measures
respondent perceptions of a whole network, by using respondents as informants about social
ties between alters as well as their own relationships. Such data may be collected via separate
questions about the outgoing ties of each actor (e.g., “Who would X go to for advice at work?”)
as in Krackhardt (1987), or by asking informant actors to fill out a matrix grid (Casciaro et al.,
1999; see Table 25.1D). Either way, a CSS task poses substantial demands on respondent time
and memory, especially for networks of even moderate size.

A CSS task yields multiple assessments{i(jk)} of each directed tie, where ai(jk) is the perception
of ordered pair (i, j) by the kth informant. Krackhardt (1987) suggests several ways of
combining these assessments to obtain a single measure for each relationship aij in a whole

network. Setting aij = ai(ji) – that is, treating each informant as the authority on her or his
outgoing ties – yields a “locally aggregated structure” essentially equivalent to the data
obtained via a typical sociometric item. An alternative “consensus structure” uses the average

assessment (where K is the total number of assessments), or a binary measure of aij

based on whether the average assessment exceeds some threshold.

Socio-Cognitive Mapping and Pile Sorts

A procedure known as socio-cognitive mapping (SCM; Cairns et al., 1985) produces a form of
cognitive social structure data that entails much lower respondent burden. It elicits respondent
perceptions of cliques or clusters. Primarily used in measuring networks among children and
adolescents, the SCM task asks respondents to report sets of people who “hang around

Page 8 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

together a lot” via free recall (see Table 25.1E). The pile-sort task used by Freeman and
Webster (1994) to measure perceived networks resembles the SCM procedure but uses
recognition methods. It provides each respondent with a randomized deck of cards containing
the names of the actors in a network, asking that the respondent sort them into mutually
exclusive piles including subsets of actors who are close to one another or who interact
frequently.

Using either the SCM instrument or the pile-sort method, each respondent's reports yield a
binary matrix with entries indicating whether the informant placed a given person in a certain
subgroup. These matrices may be combined into a consensus perception of relationships in the
whole network. Values can then be assigned to aij as a function of the number of informants
who placed two actors together in a subgroup.

Name Generator Instruments for Egocentric Networks

Suites of survey questions called “name generator” instruments elicit data on the individual
dyadic relationships and “alter” actors in a focal actor's neighborhood. They are so-called
because such instruments begin by administering one or more name generator questions that
elicit a roster of the alters within a respondent's egocentric network, thereby establishing its
boundaries. Name generators are much like sociometric items, but they almost always depend
on respondent recall, because rosters of eligible alters are typically not available in egocentric
network studies. Surveys that use name generator instruments ordinarily treat respondents as
informants who provide data about the entire egocentric network surrounding them; they
generally do not survey or interview the alters themselves. Name interpreter questions follow
the name generator(s), asking about attributes of particular alters or relationships. Such
instruments can require considerable administration time when alters or name interpreters are
numerous. Analysts later combine responses to name generators and name interpreters to
measure a wide variety of egocentric network properties.

Like sociometric items, name generator questions must specify a particular type of relationship.
Common criteria are role relations (e.g., friends, neighbors), aspects of relational form (e.g.,
closeness or frequency of contact), or specific types of resource transfer or exchange. Criteria
of the latter type, which elicit alters using a specific relational content, are especially common.
Among these is the widely used “discuss important matters” name generator (Panel A, Table
25.2) first administered in the 1985 General Social Survey (GSS) to elicit alters in a
respondent's “core” network (Marsden, 1987). Other studies use name generators tailored to
their topical content; for example, Huckfeldt and Sprague (1995) asked respondents for names
of alters whom they “talked with most about the events of the past election year.”

Page 9 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Table 25.2 Examples of name generators

A Single Name Generator (from 1985 and 2004 General Social Surveys [GSSs]; Davis et al.,
2007)
From time to time, most people discuss important matters with other people. Looking back
over the last six months, who are the people with whom you discussed matters important
to you? Just tell me their first names or initials.
IF LESS THAN FIVE NAMES MENTIONED, PROBE: Anyone else?
B Multiple Name Generator (Kogovšek et al., 2002: 14)
1 From time to time, people borrow something from other people, for instance, a piece of
equipment, or ask for help with small jobs in or around the house. Who are the people
you usually ask for this kind of help?
2 From time to time, people ask other people for advice when a major change occurs in
their life, for instance, a job change or a serious accident. Who are the people you
usually ask for advice when such a major change occurs in your life?
3 From time to time, people socialize with other people, for instance, they visit each other,
go together on a trip or to a dinner. Who are the people with whom you usually do these
things?
4 From time to time, people discuss important personal matters with other people, for
instance if they quarrel with someone close to them, when they have problems at work,
or other similar situations. Who are the people with whom you discuss personal matters
that are important to you?
5 Suppose you find yourself in a situation, when you would need a large sum of money,
but do not have it yourself at the moment, for instance, five average monthly wages
(approximately 500,000 tolars). Whom would you ask to lend you the money (a person,
not an institution such as a bank)?

To facilitate the subsequent administration of name interpreters, name generators usually ask
respondents to identify alters by first name only; egocentric designs do not need to match alter
and respondent names. A name generator may be followed by one or more probes that prompt
a respondent for additional alters. Marin (2004) demonstrates that using several such probes
can increase egocentric network size substantially, as respondents add previously forgotten
alters. Probes should be used judiciously, however, especially with behaviorally nonspecific
name generators that respondents must define for themselves. Respondents may understand
extensive probing as an indication that they are expected to cite more alters than they have
already named, leading them to alter their definition of the name generator's relational content.

Page 10 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Some name generator instruments incorporate visual interfaces to assist respondents. Kahn
and Antonucci (1980), for example, elicited social support networks by asking respondents to
place alters within concentric circles surrounding them. McCarty and Govindaramanujam
(2006) propose a dynamic visual interface in which respondents place alters in relation to one
another, thereby simplifying the collection of name interpreter data.

Two or more name generators may be used to delimit an egocentric network. Fischer (1982a),
for example, elicited social support networks using nine name generators; van der Poel (1993)
recommends sets of three and five name generators for measuring personal support networks.
The multiple-generator social support instrument in section B of Table 25.2 includes name
generators for minor instrumental aid, advice, socializing, confiding about personal matters and
major instrumental aid. Studies using multiple name generators should be mindful of possible
order effects on the number of alters given in response to particular questions (Pustejovsky and
Spillane, 2009).

As in whole-network measurement, some name generator instruments ask respondents to

name a specific number of alters. Laumann (1973) asked for three “best friends,” for example,
while Huckfeldt and Sprague (1995) asked for three political discussants, and Wellman (1979)
elicited six persons to whom a respondent felt closest. The “important matters” name generator
(Table 25.2) does not limit the number of alters but probes for more only when a respondent
names fewer than five. Many other studies (e.g., Fischer, 1982a) impose no limitations on the
number of alters.

After name generators ascertain the boundaries of a respondent's egocentric network, follow-
up name interpreter questions ask for information about its form and content. Because of the
name interpreters, egocentric network data collection poses more demands on respondents
than do whole-network instruments.

Three common types of name interpreter items are exemplified in Table 25.3. The first section
contains questions requesting proxy reports about attributes of the alters in a network, such as
race/ethnicity or age. The second section has questions about properties of ego-alter ties such
as emotional closeness, conflict/discomfort, duration of acquaintance, or frequency of contact.
Such name interpreters may ask whether a tie includes a particular strand of content of special
interest to a study. For example, the National Social Life, Health and Aging Project (NSHAP)
asked about the likelihood that respondents would discuss health with each alter (Cornwell et
al., 2009), while a 1984 South Bend, Indiana, election study asked about the frequency of
discussing politics with alters (Huckfeldt and Sprague, 1995). Finally, name interpreters may
ask about relationships among the alters themselves (the third section of Table 25.3), in order

Page 11 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

to measure egocentric network density and other aspects of egocentric network structure.

Table 25.3 Examples of name interpreters

A Name Interpreters for Alter Characteristics (from 1985 and 2004 GSSs; Davis et al., 2007)
1 Is (NAME) Asian, Black, Hispanic, White, or something else?
ASK FOR EACH NAME
2 How old is (NAME)?
PROBE: Your best guess.
ASK FOR EACH NAME
B Name Interpreters for Properties of Ego-Alter Ties (Kogovšek et al., 2002: 14–15)
1 How close do you feel to this person? Please describe how close you feel on a scale
from 1 to 5, where 1 means not close and 5 means very close.
2 How often does this person upset you?
[Responses are often, sometimes, rarely, never]
C Name Interpreters for Egocentric Network Structure (from 1985 and 2004 GSSs; Davis et
al., 2007)
Please think about the relations between the people you just mentioned. Some of them
may be total strangers in that they wouldn't recognize one another if they bumped into
each other on the street. Others may be especially close, as close or closer to each other
as they are to you.
First, think about (NAME 1) and (NAME 2).
A Are (NAME 1) and (NAME 2) total strangers?
IF YES, PROCEED TO NEXT PAIR
B Are they especially close?
PROBE: As close or closer to each other as they are to you

REPEAT FOR EACH PAIR OF NAMES

Instruments may organize name interpreter questions in “alter-wise” blocks consisting of all
questions about each alter, or “question-wise” blocks that ask a given item about all alters
(Vehovar et al., 2008). Question-wise blocking (e.g., the first section of Table 25.3) asks
respondents to answer several consecutive questions having identical response alternatives, a
“battery” format shown to elicit less reliable survey data (Alwin, 2007). Vehovar et al. (2008),
however, found question-wise presentation to be superior on several data-quality grounds, with
notably lower dropout and item nonresponse rates. Because respondents know the number of
alters they cited but not the number of name interpreters in a questionnaire, the alter-wise
format allows them to better anticipate the length of the name-interpreter task. Alter-wise

Page 12 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

presentation (e.g., the second section of Table 25.3) took less administration time in the
Vehovar et al. (2008) study, perhaps because respondents can access different pieces of
information about a given alter more rapidly in this format (Kogovšek et al., 2002).

Whether administered alter-wise or question-wise, answering name interpreter items is

repetitive and time-consuming, especially if respondents have large egocentric networks. Some
studies therefore ask name interpreters about subsets of the alters elicited. Data for a subset of
alters may be sufficient if a study seeks to measure properties of a respondent's egocentric
network, as opposed to individual respondent-alter dyads. For example, White and Watkins
(2000) elicited only up to four alters with whom respondents “chatted” because their
respondents found the name interpreters tedious. The GSS name generator instrument (Burt,
1984) asks name interpreters about only the first five alters cited (few respondents name more
than five, however). Fischer (1982a) asked some name interpreter questions about a subset of
alters, those named first in response to five different name generators. Because respondents
tend to name their closer ties sooner (Burt, 1986), selecting alters based on citation order is apt
to measure name interpreter data for stronger ties. McCarty et al. (2007) and Marin and
Hampton (2007) suggest sampling alters at random. The number of alters required to measure
a given egocentric network property with adequate reliability depends on the homogeneity of
alter characteristics within respondents (Marsden, 1993; McCarty et al., 2007).

Name generator instruments have been administered as part of face-to-face interviews as in the
GSS (Marsden, 1987), telephone interviews (e.g., Kogovšek, 2006), mail questionnaires (e.g.,
Marin and Hampton, 2007), and Web-based instruments (e.g., Vehovar et al., 2008). Name
generator instruments entail some complexity: at a minimum, names of alters must be
inventoried and organized appropriately for the administration of name interpreters. Instruments
that use more than one name generator must eliminate redundant names before presenting
name interpreters, and those that administer name interpreters for a subset of alters must use
a consistent protocol to select the subset. Well-trained interviewers can assist respondents in
avoiding organizational and navigation errors in these processes, but computer-assisted
instruments – either self- or interviewer-administered – have special appeal in this setting
(Gerich and Lehner, 2006).

To the extent that respondents view the survey content as sensitive, self-administration of name
generator instruments can enhance data quality, if it promotes higher levels of disclosure
(Gerich and Lerner, 2006). Self-administration also avoids the interviewer differences in
responses to name generators documented by van Tilburg (1998) and Marsden (2003).
Respondents may, however, err when answering self-administered instruments, for example, by
entering references to plural alters or groups (e.g., “my parents”) or other replies that do not

Page 13 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

name specific alters (e.g., “don't want to respond”) in response to name generators (Lozar
Manfreda et al., 2004). Moreover, interviewer presence may provide motivation and encourage
respondents to be attentive. Matzat and Snijders (2010) report mode comparisons that raise
concern that respondents may be prone to satisficing when answering Web-based name
generator instruments in private. In any event, the visual design of self-administered
instruments warrants care: respondents may take the amount of space left after a name
generator question as an indication of the number of alters they are expected to name, for
example (Lozar Manfreda et al., 2004; Vehovar et al., 2008).

Analysts use data from name generator instruments to construct indices that measure many
different egocentric network properties (e.g., Marsden, 1987). For respondent i, the most basic
of these is egocentric network size (ni), the number of alters elicited by the name generator(s).
Name interpreter data on relationships among alters can be used to construct measures of

local density, for example, , where r(ijk measures the strength of the tie between
alters j and k in respondent i's egocentric network. Measures of network composition can be

based on name interpreter data about alters, for example, the mean level of an attribute

(where x(ij is an attribute value for alter j in i's network) or the proportion of alters in i's network
who have a given value of an attribute. Likewise, network heterogeneity measures can be
based on the variability of alter characteristics, for example, the standard deviation

or alternatives suitable for categorical measurements. Versions of indices in Burt's

(1992) “structural hole” suite and many other egocentric network measures also can be
constructed using data from name generator instruments.

Global Questions About Egocentric Network Properties

Numerous single-item survey measures ask respondents to provide summary assessments of

some egocentric network property – most often their level or volume of informal social contact.
They do not yield data on specific actor-to-actor ties. Table 25.4 provides some examples. The
first item there asks about the frequency with which a respondent socializes with a particular
type of alter (friends outside the neighborhood), while the second asks about the size of the
respondent's network of “close friends.” The third item asks that a respondent estimate his or
her total daily number of direct social contacts, while the fourth asks for an ordinal assessment
of friendship network density. The fifth item measures the presence of a confidant.

Page 14 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Table 25.4 Examples of single-item measures of egocentric social network properties

A.Frequency of Socializing with Friends (from 1974–2008 GSSs; Davis et al., 2007)
Would you use this card and tell me which answer comes closest to how often you do the
following things …
Spend a social evening with friends who live outside the neighborhood
[Responses on card: Almost every day, Once or twice a week, Several times a month,
About once a month, Several times a year, About once a year, Never]
B.Friendship Network Size (from 1998 GSS; Davis et al., 2007)
Do you have any good friends that you feel close to?
IF YES: About how many good friends do you have?
C.Typical Daily Social Contact (Fu, 2005: 173)
On an average, about how many people do you have contact with in a typical day,
including all those who you say hello, chat, talk, or discuss matters with, whether you do
it face-to-face, by telephone, by mail or on the internet and whether you personally know
the person or not? Please give your estimate and select one from the following categories
that best matches your estimate: (1) 0–4 persons, (2) 5–9 persons, (3) 10–19 persons, (4)
20–49 persons, (5) 50–99 persons, (6) over 100 persons
D.Friendship Network Density (from 1985 GSS; Davis et al., 2007)
Some people have friends who mostly know one another. Other people have friends who
don't know one another. Would you say that all of your friends know one another, most of
your friends know one another, only a few of your friends know one another, or none of
your friends know one another?
F. Availability of a Confidant (Lowenthal and Haven, 1968)
Is there anyone in particular you confide in or talk to about yourself or your problems?

Global items like these are simple to administer within sample surveys. Their formats resemble
those of other common survey questions. They are efficient, requiring comparatively little
interview time. Some display construct validity in that they have robust statistical associations
with measures of other phenomena of interest, such as individual well-being.

Multiple-Item Instruments

Position Generator

A position generator instrument (Lin et al., 2001) measures a respondent's relationships to

particular types of alters. It does not elicit ties to particular individuals. Developed within a social
capital framework, this instrument usually assesses ties to occupational positions that vary in

Page 15 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

socioeconomic standing, presuming that alters that have more prestigious occupations offer
access to more valuable social resources. This measurement strategy also could be used to
assess ties to other types of social locations, such as ethnic or religious groups.

The position generator illustrated in Table 25.5 asks respondents to indicate whether or not
they have a specified type of contact (here, kinship, friendship, or acquaintance) with anyone in
a particular socioeconomic location. Follow-up questions ask respondents who have contact
with particular locations to indicate whether the relationship is strong (kinship, friendship) or
weak (acquaintanceship). Other follow-ups can be added.

Table 25.5 Example of position generator

Among your relatives, friends, or acquaintances, are there people who have the following jobs?

a.High school teacher

b.Electrician
c. Owner of small factory/firm
d.Nurse (etc.)

FOR EACH JOB FOR WHICH RESPONDENT ANSWERS ‘YES’, ASK:

What is his/her relationship to you?

1.Relative
2.Friend
3.Acquaintance

[IF RESPONDENT KNOWS MORE THAN ONE CONTACT WHO HOLDS A GIVEN JOB, ASK
ABOUT THE FIRST CONTACT WHO COMES TO MIND]

Responses to a position generator are usually combined into summary measures of the
composition and range of the respondent's egocentric network. Three widely used summary
measures are extensity, upper reachability, and range (Lin et al., 2001). Let xij be an indicator
variable telling whether respondent i has contact with position j, and pj be the prestige of
position j. Then the extensity of respondent i's network – the number of locations contacted –

can be measured as , where J is the total number of positions included in the instrument.
The upper reachability of i's network is defined as the highest-prestige location accessed,
. Finally, the range of a respondent's network is the difference between the highest-
prestige and lowest-prestige positions found in the respondent's network, .
Other summary measures for position-generator data can be developed via sophisticated
multivariate methods (van der Gaag et al., 2008).

Page 16 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

The position generator is a relatively efficient instrument, requiring less interview time than the
name generator instruments described above. A major decision in developing a position
generator concerns the set of positions presented to respondents. Positions presented should
cover the range of variation along dimensions underlying the locations of interest in a study
(e.g., prestige or socioeconomic standing, in the case of occupational locations) and should be
relatively common positions within the population of interest. Position generator instruments
require that respondents be able to inventory their contacts of a given type, to assess whether
one or more of them is with someone who occupies a specific position (e.g., bank teller). Such
instruments become more demanding of respondents as the number of positions and the
number of follow-up questions per position increase. In 19 surveys including position
generators reported in Lin and Erickson (2008), the number of positions ranged from 6 to 40
w i t h a m e d i a n o f 1 7 . T h e s e applications involved face-to-face, telephone, and mail
administration.

Resource Generator

The resource generator (Van der Gaag and Snijders, 2005) assesses access to social
resources directly by asking respondents if they have personal contact with anyone who
possesses certain assets or capabilities. Like position generators, resource generators do not
measure individual ties. Table 25.6 presents some example resource generator items. If a
respondent has at least one contact who controls a given resource, the instrument probes for
the strength of the strongest linkage to it.

Table 25.6 Example of resource generator

Do you know anyone who …

a.Can repair a car, bike, etc.?

b.Is handy repairing household equipment?
c. Knows a lot about governmental regulations?
d.Can give a good reference when you are applying for a job? (etc.)

[Note: the definition of ‘knowing’ a person is that the respondent would know the person's name
if he or she were to encounter the person by accident on the street and that both parties could
initiate conversation with the other.]
FOR EACH ITEM TO WHICH RESPONDENT ANSWERS ‘YES’, ASK:
What is his/her relationship to you?

1.Family member

Page 17 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

2.Friend
3.Acquaintance

[IF RESPONDENT KNOWS MORE THAN ONE CONTACT FOR A GIVEN ITEM, CODE
STRONGEST RELATIONSHIP ONLY, I.E., FAMILY MEMBER IN PREFERENCE TO FRIEND,
FRIEND IN PREFERENCE TO ACQUAINTANCE]

Like the position generator, the resource generator does not enumerate alters individually or
measure network structure: it focuses on resource-related network composition. It requires less
administration time than a typical name generator instrument. Van der Gaag and Snijders
(2005) use latent trait analysis to develop measures of aspects of social capital based on
resource generator data.

Social Support Scales

A vast literature on social support includes numerous instruments that elicit reports about the
support perceived to be available or the support actually received (Wills and Shinar, 2000).
Some measures take the form of name generator instruments that associate support with
individual alters (e.g., the second section of Table 25.2). Others ask about whether a
respondent has access to anyone who could provide a given type of support, a format like that
of the resource generator; see, for example, Cohen and Hoberman's (1983) Interpersonal
Support Evaluation List. Still other social support instruments pose separate questions about
forms of support available from classes of alters, such as family, friends, and coworkers (e.g.,
Turner and Marino, 1994).

Cognitive Considerations for Network Surveys

Recent thinking about how respondents answer survey questions stresses a four-stage
cognitive model: comprehending a question, retrieving relevant information from memory,
integrating the information retrieved (perhaps adding other considerations) to develop a
judgment about an answer, and providing a response within the format given in the survey
instrument (Tourangeau et al., 2000). This section discusses some research that bears on
these processes for questions about other persons and a respondent's relationships to them.

Sociometric questions may be misunderstood – or understood in varying ways – when they ask
about diffuse, behaviorally nonspecific relationships. One ground cited by advocates of using
“specific exchanges” (e.g., confiding, socializing) rather than affective (e.g., closeness) or role-
relation (e.g., friends, neighbors) criteria to word name generators is that respondents are apt
to answer exchange questions more consistently (McAllister and Fischer, 1978). Some research

Page 18 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

nonetheless examines variations in interpretation assigned to the widely used “important

matters” name generator. Bailey and Marsden (1999) debriefed respondents using cognitive
interviewing, finding that some took it to be asking for close or frequent contacts and that most
regarded “important matters” as those having to do with family or personal life. They suggested
that interview context might influence a respondent's definition of important matters: politics
was more apt to be part of this when a series of survey items with political content preceded the
name generator; see also Bearman and Parigi (2004). Cornwell et al. (2009) placed their name
generator instrument at the beginning of the NSHAP interview to avoid such context effects.
Bearman and Parigi (2004) asked respondents to report the important matter they had most
recently discussed, finding that matters involving money and household finance were
mentioned most frequently. A fifth of their respondents were “silent,” claiming that they had not
discussed anything important with anyone during the preceding six months.

“Friend” is among the more common role-relation criteria used in sociometric questions. Studies
using a variety of methods document wide contextual variability – notably by gender and class
– in definitions and behaviors associated with “friendship,” however (Adams and Allan, 1998).
Fischer (1982b) explored how a sample of Californians identified “friends,” finding considerable
ambiguity: they used it rather unspecifically to cover ties having no other label – often long-
duration, same-age, nonfamily contacts with whom they socialize.

Sociometric questions relying on free recall call attention to the way in which respondents
organize their memories for persons, which shapes the accessibility of information at the
retrieval stage. Studies of person memory may also suggest ways of wording questions or
probing answers to encourage accurate reporting of associates. Brewer (1995) conducted
several studies revealing that social structural factors organize memories for persons. Subjects
recalling those enrolled in their graduate program, for example, tended to give names in
clusters corresponding to entering cohorts. More generally, recall of persons corresponds to
their perceived social proximity (Brewer et al., 2005): alters perceived to interact frequently tend
to be remembered close together.

Other research by Brewer (2000) documents pervasive forgetting of associates in free recall
tasks; this appears to be more severe for weaker ties. Bell et al. (2007) reported greater
forgetting for less specific relationships (“friends” compared with sex or drug use partners), for
less salient ties (drug use versus sexual partners), and longer reference periods. The above-
mentioned research on memory organization suggests that administering probes and
reminders to call a respondent's attention to relevant contexts, or to contacts proximate to those
already named, may reduce forgetting. Brewer (2000) also suggests that instruments should
include several name generators to provide respondents with additional opportunities to name

Page 19 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

alters.

When developing name interpreter items that ask respondents to make proxy reports about
characteristics of alters, researchers should be mindful of differences between memories for
self and others (Sudman et al., 1994). People learn data about others via observation or
communication rather than experience. Memories about others may be less elaborate and
accessible and less organized into summary judgments than those about oneself.
Respondents may need to estimate rather than retrieve data about others, and they often do so
by anchoring a proxy report on their own attitude or behavior (Sudman et al., 1996). An
advantage of proxy reports is that social desirability pressures may be weaker for reports about
others than those about respondents themselves.

Apart from memory structures for persons are memories for relationships among them.
Freeman (1992) shows that people tend to impose transitivity on relationships, recalling that
others are connected when in fact they are not. This is consistent with Brewer's conclusion that
group affiliations organize person memories; nuances of within-group dyadic relationships may
not be encoded precisely, however.

It is not difficult to develop social network items that make considerable demands on
respondent cognitive capacity. Consider global items like those shown in Table 25.4. The
second item on “close friends” requires that respondents first define both “friend” and “close,”
and then enumerate or estimate their number of such contacts. Perhaps as a result, some
comparisons between responses to global items and network data assembled using other
instruments suggest that global questions have limited reliability. Sudman (1985) showed that
global estimates of network size exhibit considerable response variance by comparison to more
time-consuming network measurements based on recognition or free recall. The “ersatz”
network density item (section D in Table 25.4) makes, if anything, even stronger demands on
memory and judgment. Burt (1987) shows that answers to it are only weakly associated with a
local density measure constructed using data from a name generator instrument.

The judgment tasks involved in answering name interpreter questions about individual alters
and relationships are simpler than those posed by global items, though much more numerous.
Judgment tasks involved in answering items like those posed in position generator or resource
generator instruments can be simplified by asking – as the examples in Tables 25.5 and 25.6
do – whether anyone in an egocentric network has the attribute or resource in question, rather
than how many alters have it.

Once a respondent reaches a judgment, she or he must format it to conform with the response

Page 20 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

categories offered by a survey question. When possible, instruments should avoid presenting
response categories involving “vague quantifiers” (Bradburn and Miles, 1979), such as “rarely,”
“often,” or “some,” in favor of numerical reports or responses that involve widely understood
units of measurement (e.g., “daily” or “at least monthly”). While respondents may experience
difficulty in reaching precise judgments within the latter frameworks, using such categories
reduces measurement errors stemming from variation across respondents in classifying
judgments into vaguely quantified categories. Alwin (2007) concludes that the reliability of
reports about past behavior is lower when items use vague quantifiers.

Data Quality

Numerous sources can produce errors of measurement in survey data about social networks.
For example, errors in self-reported data about relationships can arise because of respondent
memory limitations, or because respondents seek to present themselves favorably when
answering – incorrectly claiming ties to certain alters, while omitting contacts with others.
Researchers may pose questions that correspond imperfectly with the concepts they seek to
measure, or respondents may interpret them differently than intended. Different interviewers
may administer survey items (such as name generators) in varying ways that contribute errors.

Some therefore view survey network data with skepticism. Survey designs lend themselves,
however, to obtaining the repeated measures that systematic investigations of measurement
quality require. Numerous studies examine aspects of data quality for survey-based social
network measures. Many of these offer convincing evidence that survey responses can reliably
and validly reflect underlying social network phenomena. Such studies, however, assess
quality using different standards (e.g., validity, reliability, item nonresponse), focus on different
objects of measurement (e.g., dyadic ties, characteristics of alters, egocentric network
properties), examine substantively different network ties, and measure quality using different
indices and metrics. As for survey data more generally (Alwin, 2007), quality assessments for
network data are population-specific, and the findings of any given investigation are therefore
suggestive rather than definitive.

Available data quality studies do not yield a single or unambiguous verdict about the quality of
survey measures of networks. Some measurements appear highly valid and reliable, while
others are less so. We call attention here to some influential lines of work in this area but do
not attempt either to exhaustively cover or to synthesize all relevant methodological research;
for further discussion, see Marsden (1990, 2005). We first discuss studies that assess data
quality for measures of respondent-alter relationships (sociometric items and name generators).
Subsequently, we cover research that examines name interpreter items and measures of

Page 21 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

network composition.

Accuracy and Validity Studies for Relational Items

Important and influential “informant accuracy” studies summarized in Bernard et al. (BKS;
1981) examined the accuracy or validity of reported communication ties in whole networks.
These studies compared reports about relationships based on sociometric items with “gold
standard” measurements of behavior obtained using diaries or logs or systematic observation.
Across several small populations surveyed using varying measurement methods (e.g., rankings
versus ratings), the survey reports and the “behavioral” measurements of communication
exhibited moderate agreement, at most. These findings posed serious questions about the
quality of “cognitive” reports on social ties obtained using surveys, although we note in passing
that diaries and observations of behavior also may include measurement errors. A similarly
designed classroom study (Gest et al., 2003) drew more encouraging conclusions about the
quality of survey data obtained using the SCM.

The BKS studies stimulated numerous reanalyses and follow-up studies. Kashy and Kenny
(1990) dissected the overall correspondence between survey reports and behavioral data into
two “individual” components involving actor-level tendencies to communicate with others and a
“dyadic” component involving their tendencies to communicate with particular others. Their data
analyses revealed a relatively close actor-level correspondence between survey citations
received and overall observed interaction levels, and moderate dyadic-level correlations. Survey
citations made and observed interaction levels corresponded quite poorly, however, a
conclusion later echoed by Feld and Carter's (2002) finding of “expansiveness bias” – wide
actor-level variation in reports about outgoing relationships – in survey network data. These
studies indicate that survey data may measure some network properties more validly than
others.

A separate line of work by Freeman and Romney (1987; see also Freeman et al., 1987) argued
that measurement errors in survey reports of past social interactions are not random but
instead tend to be biased toward long-term patterns. These studies obtained two-mode survey
data in which respondents reported whether others had been present at a recent meeting.
Analyses then compared those reports to attendance records – both for the particular meeting
in question and for a series of meetings held over a longer period of time. They found that
persons falsely reported as being present at the meeting in question were apt to attend most
meetings; those incorrectly reported as missing the recent meeting tended to be infrequent
attenders over the longer run. These findings suggest that survey respondents can report
stable patterns of social interaction validly, but are less capable of recalling time-specific

Page 22 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

episodes with precision. This resonates with more general findings about the cognitive
challenges respondents face when answering survey questions asking them to report event
dates (Tourangeau et al., 2000: Chapter 4).

Reliability of Relational Items

Many other studies compare two or more survey measurements of the same datum, thereby
assessing the reliability of instruments. One approach examines the reciprocation of citations,
reasoning that a conceptually undirected relationship should be reported by both actors
involved in it. Marsden (1990) inventoried several studies of reciprocation, finding rates that
ranged widely, with some indication that reciprocity is higher for closer ties. In some more
recent studies, Feld and Carter (2002) reported a reciprocation rate of about 58 percent for
college students who were questioned about who they spent time with, while White and
Watkins (2000) reported reciprocation of only 20 percent for reports about informal family
planning discussions (“chats”) by rural Kenyan women. For a high-risk population (sex workers,
intravenous drug users, and their sexual partners) adams and Moody (2007) reported much
higher reciprocation levels: 85 percent for sexual relationships, 72 percent for drug-sharing,
and 79 percent for “social” ties. These rates fell only slightly when calculations took timing into
account, counting reported citations as reciprocated only when they referred to overlapping
time intervals. Gest et al. (2003) indicate that observed interaction is higher in reciprocated
dyads, thereby offering some evidence that reciprocated citations may have higher validity.

In the data studied by adams and Moody (2007), respondents also made CSS-like reports
about contacts of their contacts. Reports about relationships among contacts of contacts were
corroborated relatively often by self-reports from participants in those relationships, though at
rates lower than the above-cited reciprocation levels. In particular, observer reports were less
often consistent with self-reports about sexual relationships, which ordinarily take place in
settings not open to observation by others.

Test-retest studies compare two or more measurements made using the same instrument on
different occasions. Test-retest correlations reflect some combination of stability in the
phenomenon under study and reliability in a measuring instrument, so they are not
unambiguous indicators of data quality. For survey network data, test-retest assessments can
be made at different levels. Some studies examine the percentage of stability/turnover in
citations of individual alters. For example, Morgan et al. (1997) showed that about 55 percent of
the alters elicited by a name generator were named again after two-month periods. In White
and Watkins's (2000) study covering a two-year interval, respondents re-named only 18 percent
of the partners with whom they first reported “chatting” about family planning. Bignami-Van

Page 23 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Assche (2005) gives an even lower figure (about 10 percent) for similar reports in Malawi that
were separated by a three-year interval. Less than 30 percent of alters were elicited on both of
two occasions separated by a much shorter period (about 10 days). Alters having close ties to a
respondent appear much more apt to be reported repeatedly across occasions than those
having weaker relationships. For reviews of other evidence on the repeated citation of alters
across waves of panel studies, see Marsden (1990) and Brewer (2000).

Instruments may measure properties of respondents’ egocentric net works reliably across
occasions of measurement despite substantial turnover in the individual alters that respondents
name, if actor-level measures of network form and composition remain stable. In one study,
Morgan et al. (1997) obtained repeated measures of egocentric network size and the percent of
family members in a respondent's egocentric network over two-month intervals, reporting
between-wave correlations above 0.6. In another, Bignami-Van Assche's (2005) name
generators elicited similar numbers of alters when readministered after a short time interval,
though respondents often did not name the same specific alters.

Several data quality studies estimate the reliability of survey network measures using a
multitrait-multimethod (MTMM) approach. Such studies repeatedly measure several
relationships using different methods and are thereby able to assess both the reliability of
measurements and the extent to which reliable variance is attributable to differences in
methods. For eight whole-network classroom studies, Ferligoj and Hlebec (1999) reported
relatively high reliability levels (above 0.85) for four social support measures. Their measures of
emotional and informational support were somewhat more reliable than those of informational
support and companionship. While measures using binary response scales appeared
somewhat less reliable than those involving ordinal scales, method-related variance in true
scores was only modest (see also Hlebec and Ferligoj, 2002).

Name Interpreter Data: Proxy Reporting and Network Composition Measures

Responses to name interpreters provide the content for most egocentric network measures
constructed using data elicited by name generator instruments. Relevant data quality studies
examine responses to name interpreters at the alter level, and they assess indices and scales
based on those data.

Proxy Reporting

When they answer name interpreters about alter characteristics, survey respondents make
“proxy” reports about alters. The survey research literature includes numerous studies
examining the quality of such reports (e.g., Moore, 1988), mostly for respondents reporting

Page 24 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

about their spouses or other household members. Alwin (2007) assessed the reliability of
spousal proxy reports about socioeconomic status, finding them to be relatively reliable –
though less so than self-reports. Sudman et al. (1994) reason that the quality of proxy reports
should rise with respondent-alter interaction.

Many studies of proxy reports obtain self-reports directly from alters, and then compare them
with an original respondent's proxy report. If we regard the alter as the authority on his or her
characteristics, such a design estimates the validity of the proxy measure. Some social network
studies use such designs to assess the quality of proxy reports about friends or other
nonhousehold alters. A common finding is that respondents can report observable data about
alters – age, sex, household possessions, number of children – reasonably well; proxy answers
about less observable features such as political party affiliation or contraceptive use are of
lower quality (Laumann, 1969; White and Watkins, 2000), and often biased toward the proxy
respondent's own value on such measures. See Marsden (1990) for further discussion.

Reliability of Network Composition Measures

As noted earlier, many measures that describe egocentric networks are within-respondent
means of data obtained via name interpreter items; for example, egocentric network density can
be expressed as the mean strength of tie between pairs of alters in a respondent's network.
Marsden (1993) assesses the reliability of such measures using methods from generalizability
theory. Their reliability rises with the number and homogeneity of measures on different alters.
Many – but not all – such properties can be reliably measured using name generator
instruments that elicit five or fewer alters.

Some other studies estimate the reliability of such measures using MTMM designs. Kogovšek
and Ferligoj (2005), for example, report reliability coefficients above 0.8 for average frequency
of contact with, and average closeness to, alters in an egocentric network. They found that
measures with behavioral rather than emotional content are more reliable. In interviewer-
administered modes, “alter-wise” presentation of name interpreters yielded more reliable
measures than did “question-wise” presentation (Kogovšek et al., 2002; Kogovšek and Ferligoj,
2005). The question-wise approach appears to be more reliable in the Internet mode, however
(Coromina and Coenders, 2006; Vehovar et al., 2008). Kogovšek and Ferligoj (2002) find that
the reliability of composition measures is higher for “core” networks composed of strong ties
than for “extended” ones that also include weaker relationships.

Another set of MTMM studies exploits the hierarchical structure of data from name generator
instruments, in which observations on alters are nested within respondents. This allows
estimation of reliability coefficients for alters within respondents and for between-respondent

Page 25 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

differences. Coromina and Coenders (2006) report reliability coefficients above 0.8 at both
levels for advice, collaboration, information exchange, and socializing among members of
research groups. Reports about collaboration were most reliable, those about socializing least
so. Coromina et al. (2004) report similar findings for common measures of tie strength, such as
closeness and frequency of contact.

Human Subject Protections and Network Surveys

Surveys that collect data on social networks must comply with all laws, regulations, and norms
that govern the conduct of survey research. Among these are obtaining voluntary, informed
consent from respondents, minimizing risks to them, and protecting their confidentiality after
data are assembled (see Citro, 2010) when disseminating research reports or archiving data
sets. Special vigilance is warranted when such surveys ask respondents for information about
their relationships involving sensitive behaviors such as sexual activity or drug use. See
Klovdahl (2005) and Woodhouse et al. (1995) for discussion of steps that may be taken under
such circumstances.

Network surveys often collect information about third parties – the alters or associates
connected to respondents. Some regard such third parties as research participants, contending
that researchers must locate and seek informed consent from them as “secondary subjects,”
though this view is debated (see Klovdahl, 2005; Morris, 2004: 3). Without question, however,
researchers are obligated to protect secondary subjects against harms arising from disclosure
of research data and to ensure that any risks to them are minimal and outweighed by potential
research benefits. Woodhouse et al. (1995) suggest that investigators may have more extensive
responsibilities to such third parties, should the research discover that they are at risk because
of the behavior of primary subjects.

Conclusion

Survey methods are, and seem apt to remain, a leading approach to collecting social network
data. Like other research methods, they have drawbacks. Assembling survey data can be both
expensive and time-consuming, especially for large samples or populations. Much
methodological research examines possible measurement errors in survey network data, but
they can also include error attributable to factors such as nonresponse or interviewer
differences.

All research data include errors, however, and such legitimate concerns about the survey
approach should not blind us to its strengths, many of which have to do with its flexibility.
Researchers control the definition of network boundaries when they conduct surveys. Survey

Page 26 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

methods can elicit all strands of relationships, not only those recorded within a specific medium
such as electronic mail. Survey researchers can measure the aspects of relationships that are
of conceptual interest in a study – rather than (for example) relying on those tracked by a
record-keeping system – thereby improving validity. Survey data are collected under relatively
standardized conditions. Surveys often require only modest time commitments from
participants, by comparison with the demands made by some alternative methods such as
diaries.

Looking forward, we can anticipate continued attention to assessing and reducing different
forms of error in survey data. As well, investigations that seek to develop more efficient
instruments and innovations in modes of collecting survey data seem likely. Internet surveys
offer many prospective gains – including substantial cost and time savings, respondent
convenience, and the possibility of using new and different visual interfaces. These innovative
methods may, however, hold consequences for data quality, investigation of which has only
begun for survey data about social networks.

Acknowledgments

For helpful comments, I am grateful to Peter Carrington, Anuška Ferligoj, and Sameer
Srivastava.

Notes

1 This view of the relationship between whole-network and egocentric data assumes, however,
that the egocentric networks for all focal actors of interest lie within whatever boundary is
established for the corresponding whole network. In practice, many egocentric studies are
conducted in large, open populations where this assumption may be untenable.

2 Obtaining egocentric network data on respondents in conventional sample surveys resembles

this design.

3 We omit discussion of network sampling methods such as respondent-driven sampling

(Heckathorn, 1997), which use social connections among elements of rare or hidden
populations to draw samples of actors in those populations, rather than to sample whole or
partial networks per se.

References

Adams, J. and Moody, J. (2007) ‘To tell the truth: Measuring concordance in multiply reported
network data’, Social Networks, 29 (1): 44–58. http://dx.doi.org/10.1016/j.socnet.2005.11.009

Page 27 of 36 The SAGE Handbook of Social Network Analysis

 SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Edited by: Adams, R.G. and Allan, G. (eds) (1998) Placing Friendship in Context. New York:
Cambridge University Press.

Alwin, D.F. (2007) Margins of Error: A Study in the Reliability of Survey Measurement.
Hoboken, NJ: Wiley Interscience. http://dx.doi.org/10.1002/9780470146316

Bailey, S. a n d Marsden, P.V. (1999) ‘Interpretation and interview context: Examining the
General Social Survey name generator using cognitive methods’, Social Networks, 21(3): 287–
309. http://dx.doi.org/10.1016/S0378-8733%2899%2900013-1

Bearman, P. and Parigi, P. (2004) ‘Cloning headless frogs and other important matters:
Conversation topics and network structure’, S o c i a l F o r c e s, 83(2): 535–57.
http://dx.doi.org/10.1353/sof.2005.0001

Bell, D.C., Belli-McQueen, B. and Haider, A. (2007) ‘Partner naming and forgetting: Recall of
n e t w o r k m e m b e r s ’, Social Networks, 29(2): 279–99.
http://dx.doi.org/10.1016/j.socnet.2006.12.004

Bernard, H.R., Killworth, P.D. and Sailer, L. (1981) ‘Summary of research on informant
accuracy in social network data, and on the reverse small world problem’, Connections,
4(summer): 11–25.

Bignami-Van Assche, S. (2005) ‘Network stability in longitudinal data: A case study from rural
Malawi’, Social Networks, 27(3): 231–47. http://dx.doi.org/10.1016/j.socnet.2005.02.001

Bradburn, N.M. and Miles, C. (1979) ‘Vague quantifiers’, Public Opinion Quarterly, 43(1): 92–
101. http://dx.doi.org/10.1086/268494

Brass, D.J. (1985) ‘Men's and women's networks: A study of interaction patterns and influence
in an organization’, Academy of Management Journal, 28(2): 327–43.
http://dx.doi.org/10.2307/256204

Brewer, D.D. (1995) ‘The social structural basis of the organization of persons in memory’,
Human Nature, 6(4): 379–403. http://dx.doi.org/10.1007/BF02734207

Brewer, D.D. (2000) ‘Forgetting in the recall-based elicitation of personal and social networks’,
Social Networks, 22(1): 29–43. http://dx.doi.org/10.1016/S0378-8733%2899%2900017-9

Brewer, D.D., Rinaldi, G., Mogoutov, A. and Valente, T.W. (2005) ‘A Quantitative review of
a s s o c i a t i v e p a t t e r n s i n t h e r e c a l l o f p e r s o n s ’, J o u r n a l o f S o c i a l S t r u c t u r e6.
http://www.cmu.edu/joss/.

Burt, R.S. (1984) ‘Network items and the General Social Survey’, Social Networks, 6(4): 293–

Page 28 of 36 The SAGE Handbook of Social Network Analysis

 SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

339. http://dx.doi.org/10.1016/0378-8733%2884%2990007-8

Burt, R.S. (1986) ‘A note on sociometric order in the General Social Survey network data’,
Social Networks, 8(2): 149–74. http://dx.doi.org/10.1016/S0378-8733%2886%2980002-8

Burt, R.S. (1987) ‘A note on the General Social Survey's ersatz network density item’, Social
Networks, 9(1): 75–85. http://dx.doi.org/10.1016/0378-8733%2887%2990019-0

Burt, R.S. (1992) Structural Holes: The Social Structure of Competition. Cambridge, MA:
Harvard University Press.

Burt, R.S. (2004) ‘Structural holes and good ideas’, American Journal of Sociology, 110(2):
349–99. http://dx.doi.org/10.1086/421787

Cairns, R.B., Perrin, J.E. and Cairns, B.D. (1985) ‘Social structure and social cognition in
early adolescence: Affiliative patterns’, Journal of Early Adolescence, 5(3): 339–55.
http://dx.doi.org/10.1177/0272431685053007

Casciaro, T., Carley, K.M. and Krackhardt, D. (1999) ‘Positive affectivity and accuracy in social
n e t w o r k p e r c e p t i o n ’, M o t i v a t i o n a n d E m o t i o n, 23(4): 285–306.
http://dx.doi.org/10.1023/A:1021390826308

Citro, C.F. (2010) ‘Legal and human subjects considerations in surveys’, in Edited by: Peter
V.Marsden and James D.Wright (eds.), Handbook of Survey Research. 2nd ed.Bingley, UK:
Emerald Group Publishing. pp. 59–79.

Cohen, S. and Hoberman, H.M. (1983) ‘Positive events and social supports as buffers of life
c h a n g e s t r e s s ’, J o u r n a l o f A p p l i e d S o c i a l P s y c h o l o g y, 13(2): 99–125.
http://dx.doi.org/10.1111/j.1559-1816.1983.tb02325.x

Coleman, J.S. (1961) The Adolescent Society: The Social Life of the Teenager and Its Impact
on Education. New York: Free Press.

Coleman, J.S., Katz, E. a n d Menzel, H. (1966) Medical Innovation: A Diffusion Study.

Indianapolis: Bobbs-Merrill.

Cornwell, B.L., Schumm, P., Laumann, E.O. and Graber, J. (2009) ‘Social networks in the
NSHAP study: Rationale, measurement, and preliminary findings’, Journal of Gerontology:
Social Sciences, 64B (S1): i47–i55.

Coromina, L. and Coenders, G. (2006) ‘Reliability and validity of egocentered network data
collected via Web: A meta-analysis of multilevel multitrait multimethod studies’, Social
Networks, 28(3): 209–31. http://dx.doi.org/10.1016/j.socnet.2005.07.006

Page 29 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Coromina, L., Coenders, G. and Kogovšek, T. (2004) ‘Multilevel multitrait multimethod model:
Application to the measurement of egocentered social networks’, Metodološki zvezki, 1(2): 323–
49.

Davis, J.A., Smith, T.W. a n d Marsden, P.V. (2007) General Social Surveys: Cumulative
Codebook: 1972–2006. Chicago: NORC.

Doreian, P. and Woodard, K.L. (1992) ‘Fixed-list versus snowball selection of social networks’,
Social Science Research, 21(2): 216–33. http://dx.doi.org/10.1016/0049-089X%2892%2990016-
A

Espelage, D.L., Green, H. D. Jr. and Wasserman, S. (2007) ‘Statistical analysis of friendship
patterns and bullying behaviors among youth’, in Edited by: Philip C.Rodkin and Laura
D.Hanish (eds), Social Network Analysis and Children's Peer Relationships. San Francisco:
Jossey-Bass. pp. 61–75.

Feld, S.L. and Carter, W.C. (2002) ‘Detecting measurement bias in respondent reports of
personal networks’, Social Networks, 24(4): 365–83. http://dx.doi.org/10.1016/S0378-
8733%2802%2900013-8

Ferligoj, A. and Hlebec, V. (1999) ‘Evaluation of social network measurement instruments’,

Social Networks, 21(2): 111–30. http://dx.doi.org/10.1016/S0378-8733%2899%2900007-6

Fernandez, R.M. (1991) ‘Structural bases of leadership in intraorganizational networks’, Social

Psychology Quarterly, 54(1): 36–52. http://dx.doi.org/10.2307/2786787

Fischer, C.S. (1982a) To Dwell Among Friends: Personal Networks in Town and City. Chicago:
University of Chicago Press.

Fischer, C.S. (1982b) ‘What do we mean by “friend”: An inductive study’, Social Networks, 3(4):
287–306. http://dx.doi.org/10.1016/0378-8733%2882%2990004-1

Frank, O. (1981) ‘A survey of statistical methods for graph analysis’, i n Edited by:
SamuelLeinhardt (ed.), Sociological Methodology 1981. San Francisco: Jossey-Bass. pp. 110–
55.

Freeman, L.C. (1992) ‘Filling in the blanks: A theory of cognitive categories and the structure of
social affiliation’, Social Psychology Quarterly, 55(2): 118–27. http://dx.doi.org/10.2307/2786941

Freeman, L.C. (2004) The Development of Social Network Analysis: A Study in the Sociology of
Science. Vancouver, BC: Empirical Press.

Freeman, L.C., Romney, A.K. and Freeman, S.C. (1987) ‘Cognitive structure and informant

Page 30 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

accuracy’, American Anthropologist, 89(2): 311–25.

http://dx.doi.org/10.1525/aa.1987.89.2.02a00020

Freeman, L.C. and Romney, A.K. (1987) ‘Words, deeds and social structure: A preliminary
study of the reliability of informants’, Human Organization, 46(4): 330–34.

Freeman, L.C. and Webster, C.M. (1994) ‘Interpersonal proximity in social and cognitive
space’, Social Cognition, 12(3): 223–47. http://dx.doi.org/10.1521/soco.1994.12.3.223

Fu, Y.-C. (2005) ‘Measuring personal networks with daily contacts: A single-item survey
question and the contact diary’, S o c i a l N e t w o r k s, 27(3): 169–86.
http://dx.doi.org/10.1016/j.socnet.2005.01.008

Gerich, J. and Lerner, R. (2006) ‘Collection of ego-centered network data with computer-
assisted interviews’, Methodology, 2(1): 7–15.

Gest, S.D., Farmer, T.W., Cairns, B.D. and Xie, H. (2003) ‘Identifying children's peer social
networks in school classrooms: Links between peer reports and observed interactions’, Social
Development, 12(4): 513–529. http://dx.doi.org/10.1111/1467-9507.00246

Hansell, S. (1984) ‘Cooperative groups, weak ties, and the integration of peer friendships’,
Social Psychology Quarterly, 47(4): 316–28. http://dx.doi.org/10.2307/3033634

Heckathorn, D.D. (1997) ‘Respondent-driven sampling: A new approach to the study of hidden
populations’, Social Problems, 44(2): 174–99. http://dx.doi.org/10.2307/3096941

Hlebec, V. and Ferligoj, A. (2002) ‘Reliability of social network measurement instruments’,

Field Methods, 14(3): 288–306. http://dx.doi.org/10.1177/15222X014003003

Holland, P.W. and Leinhardt, S. (1973) ‘The structural implications of measurement error in
sociometry’, Journal o f M a t h e m a t i c a l S o c i o l o g y, 3(1): 85–111.
http://dx.doi.org/10.1080/0022250X.1973.9989825

Huckfeldt, R. and Sprague, J. (1995) Citizens, Politics, and Social Communication: Information
and Influence in an Election Campaign. New York: Cambridge University Press.
http://dx.doi.org/10.1017/CBO9780511664113

Johnson, J.C. and Orbach, M.K. (2002) ‘Perceiving the political landscape: Ego biases in
cognitive political networks’, Social Networks, 24(3): 291–310. http://dx.doi.org/10.1016/S0378-
8733%2802%2900006-0

Kahn, R.L. and Antonucci, T.C. (1980) ‘Convoys over the life course: Attachment, roles and
social support’, in Edited by: Paul B.Baltes and Orville G.Brim (eds), Life Span Development
and Behavior, vol. 3. San Diego: Academic Press. pp. 253–86.

Page 31 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Kashy, D.A. and Kenny, D.A. (1990) ‘Do you know whom you were with a week ago Friday? A
re-analysis of the Bernard, Killworth, and Sailer studies’, Social Psychology, Quarterly, 53(1):
55–61. http://dx.doi.org/10.2307/2786869

Keating, N.L., Ayanian, J.Z., Cleary, P.D. a n d Marsden, P.V. (2007) ‘Factors affecting
influential discussions among physicians: A social network analysis of a primary care practice’,
Journal of General Internal Medicine, 22(6): 794–98. http://dx.doi.org/10.1007/s11606-007-0190-
8

Kirke, D.M. (1996) ‘Collecting peer data and delineating peer networks in a complete network’,
Social Networks, 18(4): 333–46. http://dx.doi.org/10.1016/0378-8733%2895%2900280-4

Klovdahl, A.S. (2005) ‘Social network research and human subjects protection: Towards more
effective infectious disease control’, S o c i a l N e t w o r k s, 27(2): 119–37.
http://dx.doi.org/10.1016/j.socnet.2005.01.006

Klovdahl, A.S., Potterat, J.J., Woodhouse, D.E., Muth, J.B., Muth, S.Q. and Darrow, W.W.
(1994) ‘Social networks and infectious disease: The Colorado Springs study’, Social Science
and Medicine, 38(1): 79–88. http://dx.doi.org/10.1016/0277-9536%2894%2990302-6

Kogovšek, T. (2006) ‘Reliability and validity of measuring social support networks by Web and
telephone’, Metodološki zvezki, 3(2): 239–52.

Kogovšek, T. a n d Ferligoj, A. (2002) ‘The quality of measurement of personal support

networks’, Quality and Quantity, 38(5): 517–32.

Kogovšek, T. and Ferligoj, A. (2005) ‘Effects on reliability and validity of egocentered network
measurements’, Social Networks, 27(3): 205–29.

Kogovšek, T., Ferligoj, A., Coenders, G. and Saris, W.E. (2002) ‘Estimating the reliability and
validity of personal support measures: Full information ML estimation with planned incomplete
data’, Social Networks, 24(1): 1–20.

Krackhardt, D. (1987) ‘Cognitive social structures’, Social Networks, 9(2): 109–34.

http://dx.doi.org/10.1016/0378-8733%2887%2990009-8

Laumann, E.O. (1969) ‘Friends of urban men: An assessment of accuracy in reporting their
socioeconomic attributes, mutual choice, and attitude agreement’, Sociometry, 32(1): 54–69.
http://dx.doi.org/10.2307/2786634

Laumann, E.O. (1973) Bonds of Pluralism: The Form and Substance of Urban Social Networks.
New York: Wiley Interscience.

Page 32 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Laumann, E.O. and Knoke, D. (1987) The Organizational State: Social Choice in National
Policy Domains. Madison: University of Wisconsin Press.

Laumann, E.O., Marsden, P.V. and Prensky, D. (1989) ‘The boundary specification problem in
network analysis’, in Edited by: Linton C.Freeman, Douglas R.White, and A.Kimball Romney
(eds), Research Methods in Social Network Analysis. Fairfax, VA: George Mason University
Press. pp. 61–87.

Laumann, E.O. a n d Pappi, F.U. (1976) Networks of Collective Action: A Perspective on

Community Influence Systems. New York: Academic Press.

Lazarsfeld, P.F. and Menzel, H. (1980) ‘On the relation between individual and collective
properties’, in Edited by: AmitaiEtzioni and Edward W.Lehman (eds), A Sociological Reader
on Complex Organizations. 3rd ed.New York: Holt, Rinehart and Winston. pp. 508–21.

Lazega, E. (2001) The Collegial Phenomenon: The Social Mechanisms of Cooperation Among
P e e r s i n a C o r p o r a t e L a w P a r t n e r s h i p. N e w Y o r k: Oxford University Press.
http://dx.doi.org/10.1093/acprof:oso/9780199242726.001.0001

Liebow, E., McGrady, G., Branch, K., Vera, M., Klovdahl, A., Lovely, R., Mueller, C. and
Mann, E. (1995) ‘Eliciting social network data and ecological model-building: Focus on choice
of name generators and administration of random-walk study procedures’, Social Networks,
17(3–4): 257–72. http://dx.doi.org/10.1016/0378-8733%2895%2900266-Q

Lin, N. and Erickson, B.H. (2008) Social Capital: An International Research Program. New
York: Oxford University Press. http://dx.doi.org/10.1093/acprof:oso/9780199234387.001.0001

Lin, N., Fu, Y.-C. and Hsung, R.-M. (2001) ‘The position generator: Measurement techniques
for investigations of social capital’, in Edited by: NanLin, KarenCook, and Ronald S.Burt
(eds), Social Capital: Theory and Research. New York: Aldine de Gruyter. p p . 57–81.
http://dx.doi.org/10.1017/CBO9780511815447

Lindzey, G. a n d Borgatta, E.F. (1954) ‘Sociometric measurement’, i n Edited by:

GardnerLindzey (ed.), Handbook of Social sychology. vol. 1. Reading, MA: Addison Wesley.
pp. 405–48.

Lowenthal, M.F. and Haven, C. (1968) ‘Interaction and adaptation: Intimacy as a critical
variable’, American Sociological Review, 33(1): 20–30. http://dx.doi.org/10.2307/2092237

Lozar Manfreda, K., Vehovar, V. and Hlebec, V. (2004) ‘Collecting ego-centred network data
via the Web’, Metodološki zvezki, 1(2): 295–321.

Marin, A. (2004) ‘Are respondents more likely to list alters with certain characteristics?

Page 33 of 36 The SAGE Handbook of Social Network Analysis

 SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Implications for name generator data’, Social Networks, 26(4): 289–307.

http://dx.doi.org/10.1016/j.socnet.2004.06.001

Marin, A. and Hampton, K.N. (2007) ‘Simplifying the personal network name generator:
Alternatives to traditional multiple and single name generators’, Field Methods, 19(2): 163–93.
http://dx.doi.org/10.1177/1525822X06298588

Marsden, P.V. (1987) ‘Core discussion networks of Americans’, American Sociological Review,
52(1): 122–31. http://dx.doi.org/10.2307/2095397

Marsden, P.V. (1990) ‘Network data and measurement’, Annual Review of Sociology, 16: 435–
63. http://dx.doi.org/10.1146/annurev.so.16.080190.002251

Marsden, P.V. (1993) ‘The reliability of network density and composition measures’, Social
Networks’, 15(4): 399–421. http://dx.doi.org/10.1016/0378-8733%2893%2990014-C

Marsden, P.V. (2003) ‘Interviewer effects in measuring network size using a single name
generator’, S o c i a l N e t w o r k s, 25(1): 1–16. http://dx.doi.org/10.1016/S0378-
8733%2802%2900009-6

Marsden, P.V. (2005) ‘Recent developments in network measurement’, in Edited by: Peter
J.Carrington, JohnScott, and StanleyWasserman (eds), Models and Methods in Social
N e t w o r k A n a l y s i s. N e w Y o r k: Cambridge University Press. p p . 8–30.
http://dx.doi.org/10.1017/CBO9780511811395.002

Matzat, U. and Snijders, C. (2010) ‘Does the online collection of ego-centered network data
reduce data quality?’ S o c i a l N e t w o r k s, 32(2): 105–11.
http://dx.doi.org/10.1016/j.socnet.2009.08.002

McAllister, L. a n d Fischer, C.S. (1978) ‘A procedure for surveying personal networks’,

Sociological Methods a n d R e s e a r c h, 7(2): 131–48.
http://dx.doi.org/10.1177/004912417800700202

McCarty, C. and Govindaramanujam, S. (2006) ‘A modified elicitation of personal networks

using dynamic visualization’, Connections, 26(2): 61–69.

McCarty, C., Killworth, P.D. a n d Rennell, J. (2007) ‘Impact of methods for reducing
respondent burden on personal network structural measures’, Social Networks, 29(2): 300–315.
http://dx.doi.org/10.1016/j.socnet.2006.12.005

Moore, J.C. (1988) ‘Self-proxy response status and survey response quality: A review of the
literature’, Journal of Official Statistics, 4(2): 155–72.

Moreno, J.L. (1953) Who Shall Survive? Foundations of Sociometry, Group Psychotherapy, and

Page 34 of 36 The SAGE Handbook of Social Network Analysis

SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

Sociodrama. Beacon, NY: Beacon House.

Morgan, D.L., Neal, M.B. and Carder, P. (1997) ‘The stability of core and peripheral networks
o v e r t i m e ’, Social Networks, 19(1): 9–25. http://dx.doi.org/10.1016/S0378-
8733%2896%2900288-2

Edited by: Morris, M. (ed.) (2004) Network Epidemiology: A Handbook for Survey Design and
D a t a C o l l e c t i o n. N e w Y o r k: Oxford University Press.
http://dx.doi.org/10.1093/0199269017.001.0001

Northway, M.L. (1952) A Primer of Sociometry. Toronto: University of Toronto Press.

Provan, K.G., Huang, K. and Milward, H.B. (2009) ‘The evolution of structural embeddedness
and organizational social outcomes in a centrally governed health and human services
network’, Journal of Public Administration Research and Theory, 19(4): 873–93.
http://dx.doi.org/10.1093/jopart/mun036

Pustejovsky, J.E. and Spillane, J.P. (2009) ‘Question-order effects in social network name
generators’, Social Networks, 31(4): 221–29. http://dx.doi.org/10.1016/j.socnet.2009.06.001

Singleton, L.C. and Asher, S.R. (1977) ‘Peer preferences and social interaction among third-
grade children in an integrated school district’, Journal of Educational Psychology, 69(4): 330–
36. http://dx.doi.org/10.1037/0022-0663.69.4.330

Smyth, J.D., Dillman, D.A., Christian, L.M. and Stern, M.J. (2006) ‘Comparing check-all and
forced-choice formats in Web surveys’, Public Opinion Quarterly, 70(1): 66–77.
http://dx.doi.org/10.1093/poq/nfj007

Sudman, S. (1985) ‘Experiments in the measurement of the size of social networks’, Social
Networks, 7(2): 127–51. http://dx.doi.org/10.1016/0378-8733%2885%2990002-4

Sudman, S., Bickart, B., Blair, J. and Menon, G. (1994) ‘The effect of participation level on
reports of behavior and attitudes by proxy reporters’, in Edited by: NorbertSchwarz and
SeymourSudman (eds), Autobiographical Memory and the Validity of Retrospective Reports.
New York: Springer-Verlag. pp. 251–65. http://dx.doi.org/10.1007/978-1-4612-2624-6_17

Sudman, S., Bradburn, N.M. a n d Schwarz, N. (1996) Thinking About Answers: The
Application of Cognitive Processes to Survey Methodology. San Francisco: Jossey-Bass.

Tourangeau, R., Rips, L.J. and Rasinski, K. (2000) The Psychology of Survey Response. New
York: Cambridge University Press. http://dx.doi.org/10.1017/CBO9780511819322

Turner, R.J. a n d Marino, F. (1994) ‘Social support and social structure: A descriptive

Page 35 of 36 The SAGE Handbook of Social Network Analysis

 SAGE SAGE Research Methods
©2014 SAGE Publications, Ltd. All Rights Reserved.

epidemiology’, J o u r n a l o f H e a l t h a n d S o c i a l B e h a v i o r, 35(3): 193–212.

http://dx.doi.org/10.2307/2137276

Van der Gaag, M. a n d Snijders, T.A.B. (2005) ‘The resource generator: Social capital
q u a n t i f i c a t i o n w i t h c o n c r e t e i t e m s ’, S o c i a l N e t w o r k s, 27(1): 1–29.
http://dx.doi.org/10.1016/j.socnet.2004.10.001

Van der Gaag, M., Snijders, T.A.B. and Flap, H. (2008) ‘Position generator measures and their
relationship to other social capital measures’, in Edited by: NanLin and Bonnie H.Erickson
(eds), Social Capital: An International Research Program. New York: Oxford University Press.
pp. 28–48.

Van der Poel, M.G.M. (1993) ‘Delineating personal support networks’, Social Networks, 15(1):
49–70. http://dx.doi.org/10.1016/0378-8733%2893%2990021-C

Van Tilburg, T. (1998) ‘Interviewer effects in the measurement of personal network size’,
S o c i o l o g i c a l M e t h o d s a n d R e s e a r c h, 26(3): 300–328.
http://dx.doi.org/10.1177/0049124198026003002

Vehovar, V., Lozar Manfreda, K., Koren, G. and Hlebec, V. (2008) ‘Measuring ego-centered
social n e t w o r k s o n t h e W e b ’, Social Networks, 30(3): 213–22.
http://dx.doi.org/10.1016/j.socnet.2008.03.002

Wellman, B. (1979) ‘The community question: The intimate networks of East Yorkers’,
American Journal of Sociology, 84(5): 1201–31. http://dx.doi.org/10.1086/226906

White, K. a n d Watkins, S.C. (2000) ‘Accuracy, stability, and reciprocity in informal

conversational networks in rural Kenya’, Social Networks, 22(4): 337–55.
http://dx.doi.org/10.1016/S0378-8733%2800%2900030-7

Wills, T.A. and Shinar, O. (2000) ‘Measuring perceived and received social support’, in Edited
by: SheldonCohen, Lynn G.Underwood, and Benjamin H.Gottlieb (eds), Social Support
Measurement and Intervention. New York: Oxford University Press. pp. 86–135.

Woodhouse, D.E., Potterat, J.J., Rothenberg, R.B., Darrow, W.W., Klovdahl, A.S. a n d
Muth, S.Q. (1995) ‘Ethical and legal issues in social network research: The real and the ideal’,
in Edited by: Richard H.Needle, Susan L.Coyle, Sander G.Genser, and Robert T.Trotter
(eds), Social Networks, Drug Abuse, and HIV Transmission. Rockville, MD: National Institute on
Drug Abuse. pp. 131–43.

http://dx.doi.org/10.4135/9781446294413.n25

Page 36 of 36 The SAGE Handbook of Social Network Analysis