“Elements of Structural Equation Models (SEMs)” is a highly anticipated book by the many researchers, students, and analysts who have benefitted from Dr. Ken Bollen’s seminal 1989 book, “Structural Equations with Latent Variables.” Published February of 2026, “Elements of SEMs” closes the gap between the well-established SEMs foundations of the previous century and the many methodological advances that have taken place over the past three decades. Bollen exceeds his goals of providing a comprehensive text on SEMs that is equally useful to applied researchers and methodologists, and one that serves as a learning and lasting reference tool.
The overview chapter gives readers a wonderful taste of the book’s contents. Most importantly, it sets the record straight around the foundations, meaning, goals, advantages, vulnerabilities, and myths of SEMs.
For a free download of the overview chapter, click here: A better way to answer "why" | JMP.
The chapter material
The chapter opens with the key distinction between predictive and explanatory models, reminding us that the latter are key to testing hypothesized structures and their causal inferences. SEMs are explanatory models that, since their origins in the 1920s (Wright, 1921), have required subject-matter expertise for their specification*.
“Causal assumptions are the key input to causal inferences derived from SEMs.” (pp. 11)
This requirement sheds light on one of the many advantages of SEMs: the possibility of testing, rejecting, and retaining the most probable causal hypotheses about how the world operates. Indeed, it is no surprise that SEMs are used heavily in scientific endeavors, as they fit perfectly into the notion of abductive reasoning. The American philosopher C. S. Peirce coined the term “abductive reasoning” in the early 1900s (Peirce, 1903/1997), which refers to engaging in causal inference by pitting hypotheses against each other and retaining the most possible one. The “duck test” is perhaps the best way to illustrate this concept:

We observe outcomes of a cause (appearance, walking style, and sound of a duck) and must find the most likely explanation (it’s likely a duck). Similarly, SEMs are an engine for testing causal assumptions by translating them into statistical models that can be fit to data, quantified in terms of fit, and compared against alternatives. Ultimately, we retain the most likely causal theories until new ones come along that provide an even better understanding of the world.
Bollen’s comprehensive definition of SEMs makes clear how and why SEMs fit within the larger causal inference literature. This point is critical at a time when the potential outcomes framework (Imbens & Rubin, 2015) and Judea Pearl’s work on directed acyclic graphs (DAGs; Wright, 1921; Pearl, 2009) appear to dominate the conversation.
An effective demonstration of another key feature of SEMs is introduced in the context of measurement error. Bollen introduces a path diagram with a variable that’s assessed with some degree of imperfection. When this variable is used as a predictor, the path diagram illustrates that bias in causal effects emerges because ignoring measurement error amounts to having an omitted confounder. SEMs are known for their ability to model latent variables that account for measurement error, and thus, Bollen calls into question the ubiquitous practice of ignoring measurement error in a variety of fields and applications – even those relying on propensity scores to estimate causal effects. This demonstration should alert readers to the unfortunate consequences of imperfect measurement, and arguably, should draw readers into the rest of “Elements of SEMs.”
Myths surrounding SEMs
This overview chapter also exposes a series of myths surrounding SEMs. Bollen borrows three from a previous publication where eight myths are discussed (Bollen & Pearl, 2013) and adds a new one. I find these myths of utmost importance to debunk. Thus, I encourage readers to go beyond the four outlined in “Elements of SEMs” and review the original article. To this end, I list the myths from both sources below:
- Myth #1: SEMs aim to establish causal relations from associations alone.
- Myth #2: SEMs and regression are essentially equivalent.
- Myth #3: No causation without manipulation.
- Myth #4: SEMs are not equipped to handle nonlinear causal relationships.
- Myth #5: A potential outcome framework is more principled than SEMs.
- Myth #6: SEMs are not applicable to experiments with randomized treatments.
- Myth #7: Mediation analysis in SEMs is inherently noncausal.
- Myth #8: SEMs do not test any major part of the theory against the data.
- Myth #9: SEMs assume linear Gaussian models.
The myths’ descriptions should be enough to intrigue readers.
Closing
The chapter closes with the acknowledgement of advantages – and disadvantages – of SEMs. In the end, Bollen affirms that the advantages are mostly unique to the structural equation modeling framework, whereas the disadvantages are shared with many other techniques.
In sum, readers are sure to enjoy reading this opening chapter of “Elements of SEMs” and should have even greater enjoyment taking a deep dive into the rich history and practice of SEMs as covered in the remaining chapters of the book. SEMs encompass a large variety of techniques, ranging from path analysis and measurement models to longitudinal nonlinear models for continuous, ordinal, and count data. Few people in this world have the bird’s eye view that Bollen shares with us in his latest text. Indeed, “Elements of SEMs” reflects the growth and evolution of SEMs; it’s hard to imagine a student learning SEMs without this book, and even harder to find a more comprehensive treatment on the topic.
To learn more about SEMs, please see this on-demand webinar featuring Dr. Ken Bollen: From Latent Variables to Causal Inference | JMP.
*Bollen also clarifies that structural equation modeling has grown into a powerful, general statistical framework that can simply be used for a variety of techniques that don’t necessarily focus on causal inference. It is yet another source of confusion for those new to this topic.
References
Bollen, K. A. (2026). Elements of Structural Equation Models (SEMs). Cambridge University Press.
Bollen, K. A., & Pearl, J. (2013). Eight myths about causality and structural equation models. In Handbook of causal analysis for social research (pp. 301-328). Dordrecht: Springer Netherlands.
Imbens, G. W., & Rubin, D. B. (2015). Causal inference in statistics, social, and biomedical sciences. Cambridge university press.
Pearl, J. (2009). Causality. Cambridge University Press.
Peirce, C. S. (1997). Pragmatism as a principle and method of right thinking: The 1903 Harvard lectures on pragmatism (P. A. Turrisi, Ed.). State University of New York Press.
Wright, S. (1921). Systems of mating. I. The biometric relations between parent and offspring. Genetics, 6(2), 111.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.