The DOI matches its Crossref record; 53 of the 57 references it cites were found in a registry or an index; 3 of the 45 sources its claims cite were checked for support.
The Deductive Approach to Causal Inference
Open accessAbstract
This paper reviews concepts, principles, and tools that have led to a coherent mathematical theory that unifies the graphical, structural, and potential outcome approaches to causal inference. The theory provides solutions to a number of pending problems in causal analysis, including questions of confounding control, policy analysis, mediation, missing data, and the integration of data from diverse studies.
Overview
CoChat Summary
TL;DR
Pearl reviews how nonparametric structural causal models (SCMs) unify the graphical, structural, and potential-outcome traditions of causal inference under two principles: the law of structural counterfactuals and the law of structural independence. The paper surveys complete, polynomial-time solutions to identification, mediation, transportability, and missing-data problems.
Key findings
- Identification of interventional expressions of the form P(y | do(x), z) is decidable in polynomial time, and when estimable, the estimand can be derived in polynomial time; the algorithm is complete (Summary Result 1).
- Natural direct and indirect effects are identifiable under assumption set A (A-1 to A-4), which is strictly broader than sequential ignorability (assumption set B) because it permits identification by any means and allows auxiliary covariates rather than requiring one set W to satisfy all conditions simultaneously.
- Transportability of experimental findings from multiple environments can be determined in polynomial time when differences are encoded in selection diagrams, with a transport formula derivable in polynomial time and a complete algorithm (Summary Result 3).
- Missing-data recoverability can be determined in polynomial time when the missingness process is encoded in a causal diagram falling in areas M, S, or N of Figure 5; the MNAR territory is partitioned into recoverable and non-recoverable regions, and no model-blind estimator exists for the MNAR class.
Methods
This is a theoretical review/synthesis paper, not an empirical study; it presents concepts, definitions, theorems, and "Summary Results" drawn from the author's prior work and co-authored work (e.g., Bareinboim and Pearl on transportability and meta-transportability; Mohan and Pearl on missing data). The formal apparatus includes SCMs, d-separation, the do-calculus (three inference rules), the back-door criterion, and selection diagrams. No dataset, sample, or statistical estimation is performed; the paper is based on lectures delivered at the Atlantic Conference on Causal Influence (Boston, May 15, 2013) and the JSM Meeting (August 6, 2013).
Limitations
The paper states that the area labeled "O" in Figure 5 consists of all other problem structures and that the authors "conjecture this class to be empty" (i.e., the partition is not proven exhaustive). It also notes that the partition of MNAR territory into recoverable vs. non-recoverable models is query-dependent, and that extensions from joint/conditional probabilities to causal relationships are discussed only in cited work (Mohan and Pearl), not here. Evident from the text: as a review, it reports no new empirical validation, and several results (Summary Results 1–4) are attributed to prior publications rather than derived in full within this article.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
3 Valid · 0 Not valid · 30 Not checked · 54 Own results and interpretations
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In the non-confounding case (Figure 2(a)), NDE reduces to the mediation formula:”
§ 4 Mediation analysis > 4.2 Sufficient conditions for identifying natural effects
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Finally, Section 6 describes recent work on missing data and shows that, by viewing missing data as a causal inference task, the space of problems can be partitioned into two algorithmically recognized categories: those that permit consistent recovery from missingness and those that do not.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This implies that NIE is identifiable whenever NDE and TE are identifiable.”
§ 4 Mediation analysis > 4.1 Natural direct and indirect effects
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Since all factors on the right-hand side of the equation are estimable (e.g. by regression) from pre-interventional data, the causal effect can likewise be estimated from such data without bias.”
§ 3 Intervention, identification, and causal calculus > 3.1 The rules of do-calculus
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“The following is a set of assumptions or conditions, marked A-1 to A-4, that are sufficient for identifying both direct and indirect natural effects.”
§ 4 Mediation analysis > 4.2 Sufficient conditions for identifying natural effects
What the sources say
- Not checked — no full text of the sourcecites [20] 1992, Identifiability and exchangeability for direct and indirect effects, Epidemiology source
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on 4 assumptions not shown here.
36 more claims
Create a free account to see all 87 claims and every source verdict.
- It may state two separate claims as one. 29
- Probably not a substantial claim. 24
- Not yet classified. 1
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 3 Intervention, identification, and causal calculus
“Figure 2(b) illustrates the submodel created by the atomic intervention ; it sets the value of X to x and thus removes the influence of and on X.”
§ 1. p contains at least one arrow-emitting node that is in S, or
“Figure 1 (a) Graphical model illustrating d-separation and the back-door criterion. U terms are not shown explicitly. (b) Illustrating the intervention do(X = x).$$do(X = x).$$”
Citations
Cites
The central role of the propensity score in observational studies for causal effects
ROSENBAUM, P. R., & RUBIN, D. B. (1983) · Biometrika · doi:10.1093/biomet/70.1.41 · Cited by 31,241
Statistical Analysis With Missing Data
Lahiff, M., Little, R. J. A., & Rubin, D. B. (1989) · Journal of the American Statistical Association · doi:10.2307/2289883 · Cited by 17,265
Advances in Neural Information Processing Systems 25
Zhang, Y., Sutton, C. A., Storkey, A., & Zoubin Ghahramani, A. (2012) · Cited by 12,653
Causality: Models, Reasoning, and Inference
Pearl, J. (2009) · doi:10.1017/cbo9780511803161 · Cited by 11,653
Self-citationRUBIN, D. B. (1976) · Biometrika · doi:10.1093/biomet/63.3.581 · Cited by 9,925
A test of missing completely at random for multivariate data with missing values
Little, R. J. A. (1988) · Journal of the American Statistical Association · doi:10.1080/01621459.1988.10478722 · Cited by 8,963
Statistical Analysis with Missing Data
Little, R. J. A., & Rubin, D. B. (2002) · Wiley Series in Probability and Statistics · doi:10.1002/9781119013563 · Cited by 8,473
CAUSALITY: MODELS, REASONING AND INFERENCE
Hitchcock, C. (2001) · The Philosophical Review · doi:10.1215/00318108-110-4-639 · Cited by 4,741
Self-citationCausal Diagrams for Epidemiologic Research
Greenland, S., Pearl, J., & Robins, J. M. (1999) · Epidemiology · doi:10.1097/00001648-199901000-00008 · Cited by 3,728
Self-citationAir Pollution and Hospital Admissions for Heart Disease in Eight U.S. Counties
Schwartz, J. (1999) · Epidemiology · doi:10.1097/00001648-199901000-00005 · Cited by 3,038
Research works printed but not linked: 3
- 2006, Proceedings of the twenty-second conference on uncertainty in artificial intelligence, 437
- 1993, Comment: graphical models, causality, and intervention, Stat Sci, 8, 266
- 2002, Proceedings of the eighteenth national conference on artificial intelligence, 567
Cited by
Can Consumer-Posted Photos Serve as a Leading Indicator of Restaurant Survival? Evidence from Yelp
Zhang, M., & Luo, L. (2016) · Management Science · doi:10.1287/mnsc.2022.4359 · Cited by 193
Graphical Tools for Linear Structural Equation Modeling
Chen, B., & Pearl, J. (2014) · Defense Technical Information Center (DTIC) · doi:10.21236/ada609131 · Cited by 57
“The following model, depicted in Figure 24a, represents an “encouragement design” (Holland, 1988; Pearl, 2014b) where X represents the amount of time a student spends in an after-school remedial program, H the amount of homework a student does, and Y a student’s score on the exam.”
BackgroundMethodInfluentialSelf-citationDreyfuss, J. M., Yuchi, Y., Dong, X., Efthymiou, V., Pan, H., Simonson, D. C., Vernon, A. H., Halperin, F., Aryal, P., Konkar, A. A., et al. (2021) · Nature Communications · doi:10.1038/s41467-021-27289-2 · Cited by 39
Causality, Machine Learning, and Feature Selection: A Survey
Lamsaf, A., Carrilho, R., Neves, J. C., & Proença, H. P. M. C. (2025) · Sensors · doi:10.3390/s25082373 · Cited by 51
Can User Generated Content Predict Restaurant Survival: Deep Learning of Yelp Photos and ReviewsPreprint
Zhang, M., & Luo, L. (2016) · SSRN Electronic Journal · doi:10.2139/ssrn.3108288 · Cited by 19
Metadata
- Authors
- Pearl, J.
- Affiliations
- University of California, Los Angeles (US)
- Corresponding author
- Pearl, Judea
- Venue
- Journal of Causal Inference
- Publisher
- Walter de Gruyter GmbH
- Published
- 2014-04-05
- Volume
- 2
- Issue
- 2
- Pages
- 115-129
- ISSN
- 2193-3677 (ISSN-L)
- 2193-3685
- OpenAlex
- W2109472209
- Cited by
- 28
- References
- 57
- Citations by year
- 2025: 3 · 2023: 1 · 2022: 1 · 2021: 3 · 2019: 2 · 2018: 3 · 2017: 2 · 2016: 5 · 2015: 4 · 2014: 3
- FWCI
- 4.92
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Journal of Causal Inference (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- National Science Foundation (1302448, 1249822)
- Office of Naval Research (N00014-10-1-0933, N00014)
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Modeling and Causal Inference
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Philosophy of Biology and Scientific Explanation
- Keywords
- causal inference, probabilistic graphical models, structural equations, potential outcomes, confounding adjustment, policy analysis, mediation analysis, missing data
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-28, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-28 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: The Deductive Approach to Causal Inference. https://app.cochat.ai/papers/10.1515/jci-2014-0016
- The paper
- Pearl, J. (2014). The Deductive Approach to Causal Inference. Journal of Causal Inference. https://doi.org/10.1515/jci-2014-0016
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.