Ego Network Analysis — Personal Network Analysis
- Introduced by
- Ronald S. Burt (structural holes framework); Peter V. Marsden (egocentric measures)
- Year
- 1992 (Burt); foundational measurement formalised by Marsden 2002
- Type
- Descriptive / relational network analysis
What does it do?
Ego network analysis examines the personal network of a focal individual — the ego — by mapping their direct contacts (alters) and the ties those contacts share with one another.
Formalised through Ronald Burt's structural holes framework (1992) and Marsden's egocentric measurement approach (2002), the method produces ego-level indicators such as network size, density, constraint, and brokerage role that reveal how each individual's social position shapes their access to information, resources, and influence.
The intuition behind the method
Think of each person as sitting at the centre of a web of relationships. Some people are well-connected hubs whose contacts all know one another, forming a tight cluster. Others sit between groups that would otherwise have no contact — a position Burt calls a structural hole. Being the bridge across a structural hole gives a person unique access to diverse information and the power to broker connections. Ego network analysis makes these positional advantages and disadvantages visible and measurable from data about who each person names as a contact and whether those contacts know each other.
History
The theoretical foundation of ego network analysis was laid by Ronald Burt in Structural Holes (1992), which argued that individuals bridging disconnected groups gain informational and brokerage advantages over those embedded in dense, redundant clusters. Marsden (2002) systematised egocentric measurement, clarifying how network size, density, and centrality metrics are computed from name-generator survey data. The approach builds on earlier sociological traditions of personal network research by Wellman and others in the 1970s and 1980s, and has since been extended through software ecosystems such as egor in R that standardise the three-file data structure (egos, alters, alter-alter ties).
When to use it
Ego network analysis is appropriate when the research question concerns how an individual's local network structure — rather than the global topology of a complete network — shapes outcomes. It fits descriptive, exploratory, and relational study designs, and can handle binary, continuous, or categorical tie attributes collected via survey or observation. The method requires an alter list and alter-alter tie data for every ego; without alter-alter information, structural hole and density metrics cannot be computed. A minimum of around 10 egos is needed for any meaningful analysis, and fewer than 20 egos constrain interpretation to descriptive summaries. No normality assumption applies to the network metrics themselves, though downstream regression or correlation analyses carry their own assumptions.
Limitations and considerations
- Alter-alter tie data add survey burden; respondents may underreport or misremember connections among their contacts.
- Ego networks are bounded by the respondent's awareness — contacts outside the ego's cognitive map are invisible to the method.
- Network size is cognitively constrained (approximately 150 alters per the Dunbar number), so very large personal networks are rarely captured completely.
- Survey-based data introduce representation bias: socially marginal individuals may be systematically undersampled.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Captures structural position at the individual level without needing complete network data on the entire population.
- Burt's constraint index provides a theoretically grounded, quantitative measure of structural advantage that predicts real outcomes such as career mobility and information access.
- Flexible data collection: works with survey name generators, observed interaction logs, or archival contact records.
- Applicable across social, health, organisational, and economic research where interpersonal ties matter.
Applications
- Studying how an employee's position in an intra-organisational network predicts access to novel information and career advancement.
- Public health research examining how the personal networks of patients or community members facilitate or hinder health behaviour change.
- Innovation research identifying which scientists or entrepreneurs bridge structural holes and thus act as knowledge brokers across disciplines.
- Political sociology exploring how the network composition of voters (partisan homophily versus cross-cutting ties) relates to political opinion formation.
- Development economics assessing how farmers' personal network structure affects technology adoption and risk-sharing.
How it works
How it works
Personal network data is gathered for each ego using a name-generator instrument — a survey or observation protocol that asks the respondent to list their contacts (alters) and then assesses which pairs of those alters are connected to one another (alter-alter ties). This two-level data collection yields an ego roster and an alter-alter adjacency matrix for every respondent.
For each ego, a subgraph is constructed consisting of the ego node plus all named alters and the edges between them. This ego graph captures the local neighbourhood structure from the focal individual's vantage point and is the unit of measurement for all subsequent metrics.
Structural indicators are computed on each ego graph: network size (number of alters), density (proportion of possible alter-alter ties that are present), Burt's constraint index (how tightly the ego's contacts are interconnected, limiting access to non-redundant information), and brokerage roles (how often the ego mediates between otherwise disconnected alters). A low constraint index indicates many structural holes and greater brokerage opportunity.
Ego-level metrics are compared across individuals or correlated with outcome variables such as career advancement, information diversity, or innovation. Patterns in size and density reveal cohesion versus reach trade-offs, while the constraint index quantifies the structural advantage of spanning disconnected clusters. When fewer than 20 egos are available, descriptive statistics are the appropriate fallback.
Common pitfalls
- Collecting alter lists without alter-alter tie information makes density and constraint unmeasurable — always include the alter-alter adjacency step.
- Treating ego network density and constraint as interchangeable is an error: density measures cohesion among alters, while constraint measures how much those alters restrict the ego's access to novel information.
- Applying ego network metrics to networks with fewer than 10 nodes per ego yields unstable estimates; treat results as exploratory only.
- Ignoring representation bias in name-generator surveys — low-status or isolated individuals may produce smaller alter lists not because their networks are smaller but because recall is lower.
Frequently asked questions
What data do I need to run an ego network analysis?
You need two levels of data for every focal respondent (ego): first, a list of their contacts (alters) collected via a name-generator question; second, information on which pairs of those alters are connected to each other (alter-alter ties), typically gathered with a name-interpreter follow-up. Without the alter-alter data you can measure network size but cannot compute density, constraint, or structural holes.
What is the constraint index and why does it matter?
Burt's constraint index measures how tightly interconnected an ego's contacts are. A high constraint score means the ego's alters are all connected to each other, so the ego receives the same information from multiple redundant sources and has little brokerage power. A low constraint score indicates many structural holes — the ego bridges disconnected groups and gains informational diversity and brokerage advantage. The index is computed using Burt's formula over the ego graph adjacency matrix.
How is ego network analysis different from whole-network analysis?
Whole-network (sociocentric) analysis requires complete relational data among all members of a defined population, and computes global metrics such as betweenness centrality across the entire network. Ego network analysis requires only each respondent's local neighbourhood — alters and alter-alter ties — making it practical for large or boundary-spanning populations where complete network data are infeasible to collect. The trade-off is that ego-level metrics are locally bounded and cannot reveal global network topology.
What should I do if I have fewer than 20 egos?
With fewer than 20 egos, ego network metrics become unstable and inferential analysis is unreliable. The recommended fallback is descriptive statistics: report network size and density distributions as summaries rather than inputs to regression or comparison tests, and treat findings as exploratory pending a larger sample.
Related methods
Sources
Burt, R.S. (1992). Structural Holes: The Social Structure of Competition. Harvard University Press.
ISBN: 9780674843714Copied.Marsden, P.V. (2002). Egocentric and Sociocentric Measures of Network Centrality. Social Networks, 24(4), 407-422.
Open DOI referenceCopied.