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 Today’s 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 analysis…theories, 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

3

1

2

1 3

2

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

Krackhardt’s 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 don’t 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 Monetary—outside 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

-

1

0

0

Sue

0

-

1

1

Jim

0

0

-

0

Bob

1

0

0

-

•

•

Email Communication Ed

Sue

Jim

Bob

Ed

-

4

0

2

Sue

0

-

5

1

Jim

0

0

-

0

Bob

3

0

4

-

• •

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

-

1

0

0

Ed

0

14

50000

Sue

0

-

1

1

Sue

1

15

99000

Jim

0

0

-

0

Jim

0

12

65000

Bob

1

0

0

-

Bob

0

8

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

3

6

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 organization’s 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