Stochastic Actor-Oriented Model
Also known as: SAOM, actor-based model, stochastic actor-based model, SIENA model
The stochastic actor-oriented model (SAOM), implemented in the SIENA software, is a framework for analyzing the dynamics of social networks observed at two or more time points. It treats observed network panels as snapshots of an unobserved continuous-time process in which actors, at stochastically timed moments, evaluate their local network and decide whether to create, maintain, or drop a tie so as to improve their position according to an objective function.
Key highlights
- Explicitly models the time-ordered process of tie change, allowing causal-style claims about mechanisms rather than mere cross-sectional association.
- Co-evolution models can separate social selection (similar actors befriend) from social influence (friends become similar), the central confound in peer-effects research.
- Flexible objective function accommodates a rich library of structural and covariate effects.
- Handles structurally interdependent observations correctly, unlike actor-level regressions of network change.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use SAOM when you have panel (two or more waves) network data on a roughly stable set of actors and want to disentangle the social mechanisms driving tie change — selection, reciprocity, transitivity, homophily — and, in co-evolution models, to separate selection from social influence on behavior. It assumes actors control their own outgoing ties, that change proceeds in small steps, and that the actor set changes only modestly between waves. It is not suited to single cross-sections (use ERGM), to networks with massive composition change, to relational-event streams with fine timestamps (use relational event models), or to very large networks where simulation cost is prohibitive.
Strengths & limitations
- Explicitly models the time-ordered process of tie change, allowing causal-style claims about mechanisms rather than mere cross-sectional association.
- Co-evolution models can separate social selection (similar actors befriend) from social influence (friends become similar), the central confound in peer-effects research.
- Flexible objective function accommodates a rich library of structural and covariate effects.
- Handles structurally interdependent observations correctly, unlike actor-level regressions of network change.
- Requires panel data with a largely stable actor set; large turnover or only a single observation rules the model out.
- Assumes change occurs through small, sequential, actor-controlled single-tie steps, which may misrepresent abrupt or externally imposed restructuring.
- Estimation is simulation-intensive and can fail to converge for complex specifications or large networks.
- Identifiability of selection versus influence depends on having enough waves and adequate variation; weak data can leave them confounded.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How many waves of data do I need?
Two waves are the minimum to estimate a network-dynamics model, but separating social selection from social influence on a behavior reliably usually needs three or more waves so that the temporal ordering of network change and behavior change can be distinguished. More waves also stabilize rate and objective-function estimates.
What is the difference between SAOM and a temporal ERGM?
Both model network change between panels, but SAOM is actor-driven — actors sequentially optimize their own ties in continuous time — whereas TERGM conditions the whole next network on the previous one through tie-level statistics. SAOM offers a behavioral micro-interpretation and the co-evolution machinery; TERGM stays closer to the cross-sectional ERGM logic. The choice depends on whether an actor-decision narrative fits the substantive theory.
Can SAOM model the network and an actor behavior jointly?
Yes. The co-evolution (network-behavior) model adds a second objective function governing how actors change a behavioral variable as a function of their network neighbors, and a network objective function in which the behavior can drive tie change. Comparing these two parts is how the model separates influence from selection.
Sources
- 1.Snijders, T. A. B. (2001). The statistical evaluation of social network dynamics. Sociological Methodology, 31(1), 361–395.
- 2.Snijders, T. A. B., van de Bunt, G. G., & Steglich, C. E. G. (2010). Introduction to stochastic actor-based models for network dynamics. Social Networks, 32(1), 44–60.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Stochastic Actor-Oriented Model. ScholarGate. https://scholargate.app/sociology/stochastic-actor-oriented-model