Coding Qualitative Data for Social

Network Analysis
Danielle M. Varda, PhD
Assistant Professor, School of Public Affairs

Cameron Ward-Hunt
PhD Candidate, School of Public Affairs

Outline for Todays Talk

What is SNA?
How is social network data (typically)
collected?
How is social network data coded?
Using qualitative data for SNA
Two (maybe three) examples
Issues with social network data

WHAT IS SNA?

Social Network Analysis

Social Network Analysis (SNA) is a tool used
to gather and analyze data to explain the
degree to which network actors connect to
one another and the structural makeup of
collaborative relationships (Scott, 1991).
Allows new leverage for answering standard
social and behavioral science research
questions (Wasserman and Faust 1994)

Basic Assumptions of Network Analysis

Relationships Matter
People Influence Each Other
Ideas and materials flow through relationships
Structure of relationships have consequences

Not just composition of elements of system

that matters, but also how they are put
together (how they are embedded within a
system)

Elements of SNA

Collects data on who is connected to whom

How those connections vary and change
Focus patterns of relations
Distinct from the methods of traditional
statistics and data analysistheories, models,
and applications are expressed in terms of
relational concepts or processes.

What is a Network?
A set of nodes (or actors) along with a set of
ties of specified type that link them.

Elements of a Network: Nodes

Set of actors (nodes) connected by a set of ties
Individuals
Organizations, departments, teams
These nodes have attributes
Any description of the node
Often characterized by
groups (e.g. gender, sector)

Elements of a Network: Ties

Ties connect pairs of actors
Directed (i.e., potentially
one-directional, as in giving
advice to someone)
Undirected (as in being
physically proximate)
Dichotomous (present or
absent, as in whether two
people are friends or not)
or
Valued (measured on a
scale, as in strength of
friendship)

1
1
2
1
3

1
3

Why Study Networks?

Stop the spread of disease
How relationships influence our health
behaviors
The spread of innovative practices
Study how organizations partner to leverage
resources
Anti-terrorism
For quality improvement to improve
performance

Meaning in Nodes & Lines

SNA provides an additional way to evaluate relationships
Current Assumption = More is better.
More partners = successful collaboration (counting noses)

Alternative Assumption = Less can be more.

Not based on how many partners you have, but how they are
connected.
New
Relationship

YOU

YOU

SNA is Informed by Theories

Diffusion
Contagion: Likelihood that network members will
develop beliefs, assumptions, and attitudes that are
similar to those of others in their network
Exchange and Dependency
Resource dependency

Homophily, Proximity, and Social Support Theories

Evolutionary & Coevolutionary Theories
Ecological Approaches
Age, size dependence; technological processes, community
interdependent; organizational change

2 Different Network Approaches

Whole Network
A complete set of bounded actors
Example: All members in a tobacco coalition, all public
health departments in the country, all clients in a
health delivery network

Ego/Personal Network
Randomly sample people from a population
Ask only about their alters (no roster)
Ask a sample of patients about who the members of
their personal support network are

Unit of Analysis: Whole/Sociocentric Level

Networks
Vary in Size,
Shape, and
Composition

Measures to Describe Whole Networks

Size
Inclusiveness (all minus isolates)
Component (largest connected subset)
Connectivity reachability
Connectedness pairs of nodes mutually reachable
Density
Centralization
Symmetry
Transitivity

Krackhardts Kite Network - (Centrality)

Unit of Analysis: Subgroup Level

Subgroups
are a subset
of the graph
based on
certain
nodes or
links

Unit of Analysis: Dyads/Triads

Unit of Analysis: Individual Nodes (Ego-Centric)

HOW IS SOCIAL NETWORK DATA

(TYPICALLY) COLLECTED?
20

Data Collection &Management

1. Identify the population
Bounding the network, gaining access

2. Determine the data sources

Archival, interviews, observations, surveys

3. Collect the data

Instrument design

Identifying the Population: Bounding the Study

Extremely vexing to beginners and outsiders
Network concept would seem to argue against boundaries

Empirical research makes clear we are all connected

Even if distant links dont matter, some people in the sample will be at
the edge, no matter where we cut it

Identify a boundary

Theoretical
Affiliation (Members of; Friend of)
Defined Groups (Coalitions; Employees of an Organization; Children in a Classroom)
Stakeholders (not so clear?)
Pre-Data Collection Work Might Be Necessary

22

Step 2: Determine Data Sources

Archival data/Text Analysis
Covert Networks
Citation Networks
Meeting Minutes

Surveys (online, paper, interviews; can include

network questions as part of survey)
Observations
Data Mining (internet, emails)
23

Sampling??
Can you use a sampling method to study
complete networks? In general, the answer is
no.
Exception: Egocentric

However, whole networks are not sampled

purpose is to survey the whole network!
There may be exceptions.

Step 3: Collect the Data

Surveys are typically either:
Name Generator.
unlimited in scope: the respondent may name anyone from any
sphere of life: neighbors, kin, friends, coworkers, etc.
After obtaining a large list of names, the interviewer typically goes over
each name, asking the respondent about the nature of their relationship
with that person (what social relation) and asking about attributes of that
person (sex, race, income, etc.).

Bounded List
Pre-defined list
Entire network must be identified before data collection starts
Sometimes boundaries are clear (e.g. classrooms, organizational
departments)
Sometimes not clear; might need to implement name generator
approach first

Survey Data Collection Methods

Questionnaires.
Row-based: each questionnaire forms one row in
the adjacency matrix of the group as a whole.
Use the whole matrix analytically
Each row obtained from a different source
Each could have its own measurement
idiosyncracies

Example Survey Questions

Example Survey Questions

Example Survey Questions

WHO: Name of other
organization or group
partnership?

TIMING: How
long has the
partnership
Get specifics, e.g., dept been going?
or unit, location,
Is it ongoing vs.
contact name(s).
past work?

Also note name of the

partnership itself (if it
has one).

# ___

If ended, when
and why?

a Years

___

b Months ___
1 Ongoing
2 Ceased
When & Why?

Notes:

CONTENT: What kinds of activities does ROLES: Is

the Partnership entail?
there a lead
agency or set
Mark all that apply from response to
of agencies in
question. Do not read each category
the
below, but may use them to prompt
respondent if having difficulty answering. partnership?

RESOURCES: Is there
any dedicated funding for
the Partnership, either
within the partner
organizations or from
sources outside the
Partnership?

OUTCOME:
How successful
has it been and
why? (specific
to the individual
partnership
listed below)

Focus on type of support

(and sources for outside
support), but not on
amount of funding.
1 Conduct research
9 Tools
Develop
2 Conference
10 Training
3 Educational program
11 Tech
Assistance
4 Info Dissemination
12 Legal/Regul
Change
5 Intellectual Exchang 13 New
Technologies
6 Fund Research 14 Data Repositories
7 Standards Develop
15
Advocacy/Awareness
8 Guidelines Develop
16
Other: ___________

1 No

1 Monetary either org

1 Successful

2 Yes :
____________
____________
____

2 In-kind support only

(default)

2 Somewhat
successful

3 Monetaryoutside
source

3 Not
successful

Source(s):
4 Too early to
_____________________ tell
____________________

Adding An Ethnographic Approach

Ethnography at front end helps to
Select the right questions to ask
Word the questions appropriately
Create enough trust to get the questions
answered

Ethnography at the back end helps to

Interpret the results
Can sometimes use resps as collaborators
30

HOW IS SOCIAL NETWORK DATA

CODED?
31

Data is Entered Into an Adjacency Matrix

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

R1
R2
R3
R4
R5
R6
R7
R8
R9
R10
R11
R12
R13
R14
R15
A1
A2
A3
A4
A5
A6
A7
A8
A9
A10
A11
A12
A13
A14

1 2 3 4
R R R R
- - - 1 1 1
0
1 1
1 1
0
1 1 1
1 1 1 0
0 1 1 1
1 1 1 1
1 1 0 1
0 1 0 1
1 1 1 0
1 1 1 1
1 1 1 1
0 1 1 1
1 1 1 0
1 1 1 1
1 1 1 1
0 0 0 1
0 0 0 0
0 0 0 0
1 1 0 0
0 0 0 0
1 0 0 1
1 0 0 1
0 0 0 0
0 1 0 0
1 0 1 0
0 0 0 0
0 1 1 1
1 0 0 0

5
R
1
1
1
1
0
1
1
1
1
1
1
1
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1

6
R
1
0
1
1
1
1
0
0
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0

7
R
1
1
1
1
1
1
1
0
1
1
1
1
1
1
1
0
0
1
0
0
0
0
0
1
0
0
0
0

8
R
1
1
1
1
0
1
1
1
0
0
1
1
1
1
1
0
0
0
1
0
0
0
0
0
0
1
0
0

9
R
1
1
1
1
1
0
0
0
1
0
1
1
1
1
1
0
0
0
0
0
0
0
0
1
0
0
0
0

1
0
R
1
1
1
0
1
1
1
1
1
0
1
1
1
1
1
0
1
0
0
0
0
0
0
0
0
0
1
0

1
1
R
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
0
0
0
0
1
0
0
0
1
0
0
0
0

1
2
R
1
1
1
1
1
1
1
0
1
0
1
1
1
1
1
0
0
1
0
0
0
0
0
0
0
0
0
0

1
3
R
1
1
1
1
1
0
1
1
1
0
1
1
1
1
1
0
0
0
0
0
0
0
1
0
0
0
1
0

1
4
R
1
0
1
1
0
1
1
1
1
1
1
1
1
1
1
0
0
1
0
0
0
0
1
0
0
0
0
0

1
5
R
1
1
0
1
1
1
1
0
1
1
1
0
1
1
1
0
0
0
0
1
0
0
0
0
0
0
0
0

1
6
A
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1

1
7
A
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
1
1
1
0
1
1
1
1
1
0
1
1
1

1
8
A
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
1
1
1
1
1
1
1

1
9
A
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
1
0
1
0
0
0
1
1
1
1
1
1
1

2
0
A
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
1
1
1
0
1
1
0
1
1
1
1
1
1

2
1
A
0
0
1
0
0
0
0
0
0
0
1
0
0
0
0
1
1
0
0
1
1
1
1
0
0
0
0
1

2
2
A
0
0
0
0
0
1
0
0
0
0
1
0
0
0
0
1
0
0
0
1
1

1
0
1
1
1
1
1

2
3
A
1
0
0
0
1
0
1
0
0
0
0
0
1
0
0
1
1
1
0
1
1
1
1
1
1
0
1
1

2
4
A
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
1
0
1
0
1
1
1
1
0
0
1
1
1

2
5
A
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
1
1
1
0
1
1
1
1
0

2
6
A
0
1
0
1
0
0
0
0
0
0
0
0
0
0
0
1
1
1
0
1
1
1
1
1
1

2
7
A
0
0
0
0
0
0
1
0
0
1
0
0
0
0
0
1
1
1
1
0
1
1
0
0
1
0

2
8
A
1
0
0
0
0
1
0
0
0
0
0
0
1
0
0
1
1
1
1
1
1
0
0
1
1
0
0

0
0 1
1 1 1
1 1 0 1

2
9
A
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
1
1
0
1
1
1
1
0
0
1
1
0
0

Question: Who do
you work with?
A 1 indicates the presence
of a relationship.

A 0 represents the
absence of a relationship.

Network Logical Data Structures

Friendship
Ed

Sue

Jim

Bob

Ed

Sue

Jim

Bob

Email Communication
Ed

Sue

Jim

Bob

Ed

Sue

Jim

Bob

Individual characteristics only

half the story...RELATIONS
MATTER!
People influence each other,
ideas & material flow
Values are assigned to pairs of
actors
Hypotheses can be phrased in
terms of correlations between
relations

*2012 LINKS Center Summer SNA Workshop: Analyzing Track

Relational Data & Attribute Data

Ed

Sue

Jim

Bob

Gender

Education

Salary

Ed

Ed

14

50000

Sue

Sue

15

99000

Jim

Jim

12

65000

Bob

Bob

15000

Relational Data

Attribute Data

SNA provides the ability to combine relational data with

attribute data (e.g., homophily, heterogeneity, etc)

*2012 LINKS Center Summer SNA Workshop: Analyzing Track

Graphical representation of a digraph

USING QUALITATIVE DATA FOR SNA

- 3 EXAMPLES
36

EXAMPLE 1: Qualitative Coding

of the Barbarossa Network
Cameron Ward-Hunt
PhD Candidate
School of Public Affairs
University of Colorado Denver
37

Codeword Barbarossa
Operation Barbarossa
Surprise German
invasion of the Soviet
Union in 1940

Primary Source: Codeword Barbarossa,

complied by historian Barton Whaley
(1973)
Documents 84 sub-cases with relevant
information exchanges
38

Excerpt
49. A Warning from Tito
In mid-May, while the German divisions in conquered
Greece and Yugoslavia were hurriedly being routed through
Belgrade toward Rumania, another opportunity for a
credible disclosure existed. Vladmie Dedijer reveals in his
official biography of Tito: A senior German officer told a
Russian refugee that Hitler was preparing to attack Russia.
This information reached Tito, who sent a radiogram to
Dimitrov toward the end of May bringing it to his notice.
Dimitrov, in Moscow in his capacity as secretary-general of
the Comintern, would have immediately informed the
NKVD, if not other Soviet authorities, of such intelligence.

39

Coding Example

Josef Masin
Josef Stalin
Josef Tito
Khlopov
Konon Molody
Konstantain Umansky
Laurence Steinhardt
Leopold Trepper
Lieutenant Colonel Louis Baril
Lieutenant Commander Alwin (The Shadow) Kramer
Lieutenant-General Ivanovich Golikov
Lieutentant-General M.A. Purkayev
Lord Casey
Louis Lochner

Harry Flannery

Harry Carlson

Harold H. Tittleman

Type
Value
Message
6
Leaked
Document
9
Message
6
Message
3
Message
3
Observable
1
Observable
2

Hans Lazar

Hans Heinrich Herwarth von Bittenfield

Gustav Hilger

German Sergeant-Major Deserter

Georgi Dimitrov

Generalissimo Chiang Kai-shek

Coding of the
Information Exchange
Network
1) Extraction
2) Matrix coding

General Sikorski

47 23-May-41From Beck to Maass to Lochner to TASS to GRU and NKGB

48 22-May-41From Khlopov to GRU headquarters
49
May-41From Unk German Officer 2 to Unk russian refugee to Tito to Dimitrov to NKVD
53 1-Jun-41From Etzdorf to Lanza
51
May-41Observable in Court photographer window to Berezhkov
52
Jun-41Observable to Kelly

General S.I. Kabanov

AP channel
GRU in Berlin
Tito
Napoleonic Clue
Map clue
Counterfeit Rubles

Case Date
Narrative
46 15-May-41From Hans Lazar to Kowalewski through Pangal - to Polish govt in exile

General Georgy Zhukov

Info
Press leak

40

Coding Example

Louis Lochner
Khlopov
Josef Tito
UNK German Officer2
UNK Russian Refugee
Georgi Dimitrov
Dr. Hasso von Etzdorf
Michele Lanza
Admiral Kuznetsov
Admiral Francois Darlan

Admiral Francois Darlan

Admiral Kuznetsov

Michele Lanza

Dr. Hasso von Etzdorf

Georgi Dimitrov

Echelon
Position
Strategic
Diplomatic
Diplomatic
Tactical
Covert
Strategic
Strategic
Diplomatic
Strategic
Covert
Diplomatic
Head of State
Strategic
UNK Russian Refugee

For Attribute File

For Network Matrix
(also in binary)

Louis Lochner

Social Network Coding

Location
Russia
Portugal
Russia
France
Russia
Sweden
Russia
Germany
England
Russia
Germany
UNK German Officer2

Nationality
Soviet
Hungarian
Soviet
Soviet
Soviet
Soviet
Bulgarian
Soviet
Soviet
Soviet
Soviet

Josef Tito

Position
Commisar of the Navy
Hungarian Minister
Soviet Deputy Foreign Commisar
GRU agent
Chief of Staff, Moscow
Commanding Officer, Soviet Base Hango Peninsula
Comintern Secretary General
Chief of TASS Bureau-Berlin
Soviet Ambassador to the United Kingdom
General Secretary
Deputy Military Attache in Berlin

Khlopov

Name
Admiral Kuznetsov
Andre de Vodianer
Andrey Vyshinsky
Carlo
General Georgy Zhukov
General S.I. Kabanov
Georgi Dimitrov
Ivan Filippov
Ivan Maisky
Josef Stalin
Khlopov

24
23
23
23

41

Example Findings
RQ3: How do nations share intelligence information?
Figure 3 Barbarossa Social Network by Nationality
German

Strong international
social network =
Potential for
communication
19 nationalities
18 locations
American

Soviet
But does the potential network
translate to information shared?
Soviet (N=34), German (N=31), American (N=26), British (N=12).
42

Example Findings
RQ3: How do nations share intelligence information?
Figure 4 Barbarossa Information Network by Nationality
43.8% of all transactions
occurred between
participants of different
nationalities

13.6% shared by
diplomatic ties

Soviet

American

Brokerage Roles,
InfoNet:Nationality
2%

Coordinator

12%

68%

6%

Gatekeeper

12%

Representative
Consultant
Liaison

Conclusion
Robust percentage of sharing
outside of diplomatic channels
Different sharing patterns (i.e.
Americans versus British)

43

EXAMPLE 2: COLLABORATING FOR

IMPACT: USING SOCIAL NETWORK
ANALYSIS TO EXPLORE NONPROFIT
COMMUNITY INTERCONNECTIONS

Data
Dataset drawn from a community of nonprofit
organizations
Online website, GivingFirst, where nonprofit organizations in the
greater Metro Denver area post detailed profiles of their
organizations in order to raise funding for their organizations.
Databas

Variables we coded included:

Number of staff in the organization (including full-time, parttime, volunteer, and contractors),
Governance information (number and names of the Board of
Directors members),
Revenue information,
Mission or purpose of the nonprofit organization, and
Each organizations partnerships and affiliations

Example of Text We Coded

46

How We Coded This

47

Data
Respondent (organizations that posted profiles),
N= 362
These 362 organizations identified 2219 other
organizations as either partners or affiliates
In total, 3765 dyads (or relationships) were
generated.
Of these dyads, 3149 were identified by respondents
as collaborations and 616 as affiliations.
The data analysis was performed only on the 3149
collaborations.

UCINET used for exploratory SNA

Connectivity
Fully Connected
All nodes reachable
Most with 1 (N=1087), 2 (N=301), 3 (N=135), 4 (N=181), 5 (N=97), 6
(N=30), 7 (N=84), 8 (N=36), 9 (N=87), 10 (N=131), 11 (N=54), 12
(N=46), 13 (N=9),
Layers of connectivity

Components
One large component; 21 other components
Made up of policy areas: Behavioral Health, Courts/Offender
Programs, Dance/Theater, Environmental, Faith-Based, Health,
International Development, International Human Rights, County
Organizations, Music (Band), Parochial Schools, Prisons/Reentry,
Rotary, Spanish Arts, Sports (Soccer), Young Adults, Water, some
uncategorized because orgs not consistently servicing one area.
Not grouped by NTEE-CC categories

Key Players InDegree & OutDegree

OutDegree
American Humane Association

119

Colorado Humanities
Share Our Strength's Operation Frontline CO

85
69

AfricAid, Inc.

65

Parenting Place

55

Autism Society of Colorado

50

Street's Hope

48

Cross Community Coalition

41

ACCESS Housing
Colorado Dragon Boat Festival
InDegree

40
33

Denver Public Schools

26

University of Denver

18

Food Bank of Rockies

14

Denver Health Medical Center

13

Mile High United Way

13

Head Start

12

Colorado Nonprofit Association

11

Colorado Coalition for the Homeless

10

SafeHouse Denver, Inc

Family Tree, Inc.

9
9

Brokerage

Discussion Points
Nonprofit Communities are highly connected
Connections tend to form based policy areas,
rather than NTEE categorization
Connections are based on need (resource
dependency; access to client population) etc.

Connections within groups tend to be

Coordinating positions
Certain types of categories act more as brokers than
others

Organizational capacity seems to have something

to do with # of connections
Betweeness seems to have more to do with the
description of the clients served

EXAMPLE 3:
COLLECTING DATA FROM A
COMMUNITY COALITION TO INFORM
QUALITY IMPROVEMENT

Using SNA for CQI

Network data tell us about how people/organizations
are connected including the quantity and quality of
those connections.
Alone = hard to interpret or use in practice
Instead = Strategic Network Management (CQI process)
Identifying the ideal network.
Measuring the Network
Identifying the gap between the actual and ideal
network
Creating action steps to get closer to the idea.

Collecting Data A Hands On Approach

Who: Early Learning; Family
Support & Parent Education;
Social Emotional & Mental Health;
Health
Purpose: To identify stakeholders
and ideal system

What the Groups Produce

57

Coding the Pictures of Ideal Systems

58

ISSUES WITH SNA DATA

59

Issues with SNA Data

Response bias
Asymmetry
Missing data
Accuracy
Ethics

60

Ethical Issues

Respondents cannot be anonymous

Non-respondents are still included
Missing data can be powerful
Has the potential to be mis-used by
Management

61

Data Collection Limitations

Informant accuracy
Can people really tell you about their social networks?
Marketing researchers have found that consumers can barely
tell you what they had for lunch yesterday. Bernard, Killworth
and Sailer investigated informant accuracy systematically and
found that about 52% of what they said was wrong.
Based on the work of Freeman, Freeman and Romney, as well
D'Andrade, DeSoto, and many others, it appears that people's
recall of their interactions with others is systematically biased
toward what is normal and/or logical.

Data Collection Limitations

People also tend to remember interactions with
people who are important, while forgetting
interactions with people that are not.
Some respondents will lie to make themselves look
good, since people judge others on who they
associate with.
As with any questionnaire, there are also problems
with how people interpret the questions. What
"friend" means to one person may be very different
from what friend" means to others.

Resources

SNA Professional Organization

wwww.INSNA.org

Comprehensive List of Courses

http://socialnetworkcourses.wordpress.com/2
010/11/11/list-of-snsna-courses/

Office of Behavioral & Social Sciences Research

http://obssr.od.nih.gov/scientific_areas/meth
odology/systems_science/index.aspx

List of Recommended Readings

http://obssr.od.nih.gov/pdf/valente_recomen
_readings.pdf

UCINET
http://www.analytictech.com/ucinet/

Online SNA Text (UCINET)

http://www.faculty.ucr.edu/~hanneman/nettext/

PARTNER
(Program to Analyze, Record, and Track
Networks to Enhance Relationships)
www.partnertool.net