Exam STAM vs. Exam MAS-II: Short-Term Math Comparison

SOA Short-Term Content: STAM, FAM, and ASTAM

The SOA's short-term actuarial mathematics content has undergone significant restructuring in recent years. The original Exam STAM (Short-Term Actuarial Mathematics) covered loss models, frequency-severity distributions, aggregate loss models, credibility theory, and insurance pricing methods in a single exam. The SOA has since incorporated this material into Exam FAM (Fundamentals of Actuarial Mathematics) and Exam ASTAM (Advanced Short-Term Actuarial Mathematics), splitting the content across two levels to provide a more gradual learning progression.

The SOA's approach emphasizes mathematical derivation and theoretical understanding of loss distributions. Candidates learn to work with parametric distribution families (exponential, Pareto, Weibull, lognormal, gamma), fit distributions to data using maximum likelihood estimation and method of moments, and understand how insurance mechanisms such as deductibles, policy limits, and coinsurance modify loss distributions. Credibility theory is taught using both limited fluctuation (classical) and greatest accuracy (Buhlmann and Buhlmann-Straub) approaches. This content provides a foundation in P&C actuarial mathematics that SOA candidates need regardless of their intended specialization, because many actuarial concepts cross practice areas.

CAS Exam MAS-II: Modern Actuarial Statistics II

Exam MAS-II covers advanced statistical methods with direct applications to P&C insurance. The syllabus includes credibility theory, Bayesian estimation, generalized linear models (GLMs) for insurance pricing, loss distributions and severity modeling, and machine learning techniques including penalized regression, classification methods, and model validation. The exam is 4 hours with approximately 40 questions, requiring both computational ability and conceptual understanding of when and why to apply each method.

MAS-II represents the CAS's commitment to integrating modern data science techniques into the actuarial credential. The exam was designed to ensure that CAS-credentialed actuaries are proficient in the statistical methods used in contemporary insurance pricing, reserving, and underwriting. CAS candidates frequently report that MAS-II content directly applies to their day-to-day work, particularly in building and validating GLM-based rating plans and in applying credibility-weighted estimates to sparse data segments.

Topic Overlap and Differences

Several topics appear in both the SOA and CAS content, but the treatment differs significantly. Credibility theory is a major topic in both paths. The SOA treatment emphasizes the mathematical derivation of credibility formulas, including proofs of the Buhlmann credibility premium and the conditions under which it is optimal. The CAS treatment through MAS-II also covers credibility rigorously but places greater emphasis on Bayesian approaches to credibility and the connection between classical credibility and Bayesian estimation.

Loss distributions are another area of overlap. The SOA content focuses on fitting parametric distributions to loss data, testing goodness of fit, and calculating modified distributions under insurance mechanisms. The CAS content addresses similar material but integrates it more closely with GLM-based pricing, where the distribution of losses is a component of the model specification rather than a standalone exercise. Generalized linear models receive deeper treatment in MAS-II, where candidates learn to specify and fit GLMs for insurance pricing, select appropriate link functions and error distributions, and assess model performance using deviance and other diagnostic measures.

Philosophical Differences in Approach

The most fundamental difference between the SOA and CAS approaches to this material is philosophical. The SOA's content emphasizes classical mathematical statistics and actuarial science, teaching candidates to derive and apply established actuarial formulas in the context of loss modeling. The approach is rigorous and theoretical, with a focus on understanding why formulas work and under what conditions they are valid.

The CAS approach through MAS-II is more applied and forward-looking, emphasizing modern statistical and machine learning methods that reflect how actuarial work is actually performed in contemporary practice. The inclusion of topics like penalized regression (lasso, ridge, elastic net), classification methods, and model validation reflects the CAS's view that the traditional actuarial toolkit must be supplemented with methods from data science to remain relevant. This practical orientation means that CAS candidates may be better prepared for the technical aspects of modern pricing work, while SOA candidates may have a stronger theoretical foundation in classical loss modeling.

Difficulty Comparison

MAS-II is generally considered more challenging than the SOA's equivalent content for several reasons. The MAS-II syllabus is broad, spanning classical credibility, Bayesian statistics, GLMs, loss modeling, and machine learning in a single exam. The depth of the GLM and Bayesian content exceeds what SOA candidates encounter at the corresponding level. Pass rates for MAS-II have typically been in the 35% to 45% range, among the lowest of any preliminary-level exam in either organization. Study time averages 350 to 450 hours, reflecting the volume and complexity of the material.

The SOA's restructured content (FAM plus ASTAM) spreads the short-term material across two exam sittings, which may make it more manageable than tackling all of MAS-II in a single sitting. However, the total study time required for FAM and ASTAM combined may be comparable to MAS-II, and candidates must pass two separate exams rather than one. The trade-off between breadth per sitting and number of sittings is a practical consideration for candidates planning their exam progression.

Practical Applications

Both paths prepare candidates for actuarial work involving short-term insurance risks, but the practical applications differ in emphasis. SOA candidates with STAM/FAM/ASTAM knowledge can analyze loss distributions, estimate reserves using credibility-weighted methods, and understand how insurance policy provisions affect expected claim costs. These skills are valuable across insurance practice areas, including health insurance, where short-term loss modeling concepts apply to medical claims data.

CAS candidates with MAS-II knowledge are particularly well prepared for building and validating GLM-based pricing models, which have become the standard approach to personal and commercial lines rating in the P&C industry. They also bring stronger Bayesian estimation skills, which are useful in situations where data is sparse and prior information can improve estimates. The machine learning content provides a foundation for more advanced predictive modeling applications, such as telematics-based auto insurance pricing and automated underwriting systems.

Study Tips

For SOA candidates preparing for the short-term content, focus on mastering the parametric distribution families and their properties. Be comfortable with maximum likelihood estimation and goodness-of-fit testing. Understand how deductibles, limits, and coinsurance modify loss distributions, and practice working through these modifications for common distribution families. For credibility, work through both the limited fluctuation and Buhlmann approaches, understanding when each is appropriate.

For CAS candidates preparing for MAS-II, allocate study time carefully across the broad syllabus. Do not underestimate the Bayesian content, which is often the most challenging section for candidates without prior Bayesian statistics coursework. For GLMs, practice not only fitting models but also interpreting coefficient estimates in the context of multiplicative insurance pricing structures. For machine learning topics, focus on understanding the concepts and their insurance applications rather than memorizing algorithmic details. Practice problems and past exam questions are essential for both paths, as both exams test the ability to apply concepts to specific scenarios rather than reproduce memorized derivations.