SOA Exam SRM Study Guide (2026)
SOA Exam SRM (Statistics for Risk Modeling) is a 3.5-hour, 35-question multiple-choice exam testing the statistical and predictive modeling techniques actuaries use in risk analysis (SOA). The pass rate ran 71.2% at the January 2026 sitting, on a pass mark of 63% average correct, and has held between 67.9% and 71.2% across the last three sittings (SOA). SRM passes at a notably higher rate than most SOA preliminaries. Exam SRM sits in the SOA's preliminary exam sequence, usually after Exams P, FM, and FAM.
SRM is the point where the actuarial sequence pivots from pure mathematics to applied data science. The content overlaps heavily with what data scientists and quantitative analysts learn, just framed through an actuarial lens. If you liked the statistics and modeling courses in your degree, SRM will feel natural.
What the Exam Covers
The SOA syllabus splits SRM into five topic areas (SOA):
1. Basics of Statistical Learning (5 to 10%)
Supervised versus unsupervised learning, regression versus classification, model flexibility and the bias-variance tradeoff, training versus test error, and resampling methods (validation set, k-fold cross-validation).
2. Linear Models (40 to 50%)
The biggest block by far, and the SOA groups both classical regression and GLMs under it. Classical side: simple and multiple linear regression, assumptions and diagnostics, residual analysis, multicollinearity, interpretation of coefficients, R-squared and adjusted R-squared, plus regularized regression (ridge, lasso) and K-nearest neighbors.
The GLM side is the heart of the exam:
- Link functions (log, logit, identity)
- Distributions in the exponential family (Poisson, binomial, gamma, normal)
- Logistic regression for binary outcomes
- Poisson regression for count data
- Deviance, AIC, BIC for model comparison
- Variable selection (forward, backward, stepwise)
- Overdispersion and how to address it
Linear Models is the single most important topic area on Exam SRM. The SOA weights it at 40 to 50% of the exam, and its GLM half is how actuaries actually model insurance data. Put at least 40% of your study time here.
3. Time Series Models (10 to 15%)
Autoregressive (AR), moving average (MA), and ARIMA models, plus exponential smoothing. Stationarity, differencing, autocorrelation and partial autocorrelation functions (ACF/PACF), model identification, and forecasting.
4. Decision Trees (20 to 25%)
Classification and regression trees (CART), pruning, bagging, boosting, and random forests. This is the second-largest topic area on the current syllabus, so give it real study time.
5. Unsupervised Learning Techniques (10 to 15%)
Principal components analysis (dimensionality reduction, eigenvalues, loadings, proportion of variance explained), K-means clustering, hierarchical clustering, and methods for choosing the number of clusters. The exam tests conceptual understanding and interpretation of PCA and clustering output.
Candidates who studied machine learning in school sometimes blow through GLMs, figuring logistic regression is "easy." The exam tests GLMs at a depth that goes past most intro ML courses. You need deviance residuals, link function selection, and overdispersion diagnostics, not just the ability to fit a model.
Exam SRM Syllabus: Statistical Learning Methods
The Exam SRM syllabus is, at its core, a tour of statistical learning methods: linear regression as the foundation, generalized linear models as the workhorse, then the supervised and unsupervised techniques (decision trees, random forests, principal components analysis, cluster analysis) that fill out a modern predictive-modeling toolkit. Time series is the one block that sits apart from that framing, covering ARIMA models and forecasting.
The syllabus rewards interpretation over computation. Most questions hand you fitted output, a regression table, deviance statistics, an ACF plot, and grade whether you can read it, compare candidate models, or spot what is wrong. That is why the reference list is anchored by ISLR, a statistical learning text, rather than a classical probability text.
The SOA updates the syllabus and reading list periodically, so check SOA's Exam SRM page for the current syllabus and readings before locking in a study plan. The topic areas and weights above track the current version.
The ISLR Advantage
The primary reference text for Exam SRM is An Introduction to Statistical Learning (ISLR) by James, Witten, Hastie, and Tibshirani. The book is freely available from the authors' website. That is a real edge over other actuarial exams, where the reference texts run $100+.
ISLR is well-written and accessible. Read it carefully, work the conceptual exercises, and you will have solid coverage of most Exam SRM topics. The one gap: ISLR does not cover GLMs to the depth the SOA wants, so plan on supplementary material there.
Study Timeline (13 Weeks)
Plan for roughly 250 to 300 hours for SRM. A 13-week plan at 20 to 23 hours per week:
Weeks 1 to 3: Linear Regression
Build a thorough foundation. Do not rush it. Every GLM concept extends from linear regression. Practice reading regression output, checking assumptions, and knowing what each diagnostic is telling you.
Weeks 4 to 7: GLMs (The Core)
Spend the most time here. Work through Poisson regression, logistic regression, and gamma regression examples. Practice calculating deviance, comparing models with AIC/BIC, and spotting overdispersion. This is where the exam separates passing candidates from failing ones.
Weeks 8 to 10: Time Series, PCA, Trees, and Clustering
These topics weigh less individually but show up on every exam. Time series is the most calculation-heavy of the four. PCA and clustering lean conceptual.
Weeks 11 to 13: Practice Exam Phase
Full-length practice exams (35 questions, 3.5 hours). You get 6 minutes per question on average, which is generous next to Exams P and FM. Use the cushion to read questions carefully and double-check your reasoning.
Exam SRM gives you 6 minutes per question, a full minute more than Exam FM's five. The extra time per question reflects how interpretive the material is. Use it. Read the output tables closely before you answer.
Practice Strategy
Plan for 700 to 1,000 practice questions before exam day.
Exam SRM questions tend to hand you model output (regression tables, deviance statistics, ACF plots) and ask you to interpret it or pick the best model. That is a different game from Exams P and FM, where you crank calculations from scratch.
Work the free SRM sample questions and full practice bank, 1,028 questions with adaptive difficulty and detailed solutions covering all five topic areas.
Practice benchmarks:
- After 300 questions: scoring 45 to 55% on mixed practice
- After 600 questions: scoring 55 to 65% on practice exams
- Exam ready: scoring 65%+ consistently on full-length mocks
Common Mistakes
Rushing through GLMs. Underinvesting in GLMs and overinvesting in topics like PCA or clustering is a bad trade. GLMs are 30 to 35% of the exam.
Memorizing formulas without understanding them. SRM tests interpretation, not computation. Knowing the formula for AIC matters less than knowing when AIC points to a different model than BIC.
Ignoring model diagnostics. The exam keeps asking "what is wrong with this model?" or "which diagnostic flags a problem?" Practice reading residual plots and goodness-of-fit statistics.
Some candidates study SRM like Exam P, drilling calculations. SRM is fundamentally an interpretation exam. You need to know what the numbers mean, not just how to compute them.
Free Resources
- FreeFellow Exam SRM Practice - 1,028 free practice questions with worked solutions and a free readiness score. Adaptive difficulty and analytics are part of the paid Fellow tier.
- ISLR (free textbook) - the primary reference, freely available from the authors
- Exam SRM sample questions - all 66 SOA-published sample items with worked solutions
- SOA study note on GLMs - supplementary material covering GLMs beyond what ISLR provides
Start your Exam SRM preparation with free practice questions on FreeFellow.