Studying for the SOA Exam FAM (Loss Models Component)
The FAM exam covers probability, loss models, credibility, and risk management all in one sitting. The loss models section alone takes up roughly 35 to 40 percent of the exam, and that is where most people lose points. I have seen it enough times that I do not need to tell you it is a problem area. The L portion refers to the Loss Models material, which is what used to live inside Exam LTAM before SOA merged everything into FAM. You need to understand continuous and discrete distributions, censored and truncated data, parameter estimation through maximum likelihood and method of moments, goodness-of-fit testing with chi-squared and Kolmogorov-Smirnov, and the basic credibility frameworks like Bühlmann and Bühlmann-Straub. That is the core. Anything outside of that range shows up occasionally but rarely carries more than a couple questions. The biggest mistake I see is treating loss models like a memorization exercise. The formulas themselves are straightforward, but the questions are designed to make you pick the wrong one under time pressure. For example, you will routinely see a dataset described as left-truncated at some value d, and the question will ask you to estimate a parameter. If you apply a standard maximum likelihood formula without adjusting the likelihood function for the truncation, you will get the wrong answer and waste two minutes figuring out why. The fix is simple once it clicks: every observation above the truncation point needs to be divided by the survival function evaluated at the truncation threshold. Write that down on your first practice problem and never forget it.
A second common trap involves right-censored data mixed with exact observations in the same dataset. Students often forget that censored observations still contribute to the likelihood, just through the survival function instead of the density. I ran into this on a practice set where the dataset had twenty exact claims and ten right-censored claims at time 50. I initially built the likelihood using only the density terms and got a parameter estimate that was clearly too low. Once I included S(50) for each censored observation, the estimate shifted into a reasonable range. The workaround is to label every data point as either exact or censored before writing any equations. That single step prevents the error entirely.
How to Structure Your Study Time
Loss models typically requires about forty to fifty hours of focused study if you already have a working knowledge of probability. If your probability is shaky, budget closer to sixty hours. The earlier parts of the exam like probability theory and credibility can absorb another thirty to forty hours depending on your background. I would recommend starting with loss models first rather than last. The material builds on itself, and getting it out of the way early means you are not carrying anxiety into the credibility and probability sections. Work through Kaplan or ASM notes, then switch to Don McLeish’s lecture videos if you need more rigorous coverage. The Actex manuals are adequate but occasionally sparse on the estimation side.
Get the Full Details
Practice Problems That Actually Help
Past exam questions from the old LTAM and P exams are still relevant. The topic coverage overlaps heavily. The SOA has released sample questions for FAM that are useful too, though they are fewer in number than the older releases. I used the Altus ExamFam practice sets and the ASM practice exam bank. Doing forty to fifty loss model specific problems under timed conditions, without notes, will tell you more than any review video. If you can solve those cleanly, you are in good shape for that section. The official SOA exam website at soa.org lists the current FAM syllabus and the sample questions. You can find the syllabus document there directly. For supplementary material, the ASM Study Manuals for FAM are available through their website and major retailers. Kaplan also publishes a FAM prep course with video lectures and problem sets. Actex has a FAM manual and a separate loss models review book that some people prefer for deeper coverage of the estimation and fitting topics. The FAM exam combines four distinct topic areas into one sitting, and that creates a real bottleneck. If you are strong in probability but weak in loss models, the exam will still punish you because the sections are weighted together. There is no way to compensate for a weak area with strength in another. The format also means you cannot predict with certainty which topics will carry more weight in any given administration, since SOA does not release exact weights in advance. Plan your study accordingly and do not gamble on any single section being light.
Another limitation is the calculator dependency. The exam allows only the TI-30X S and the BA II Plus Professional. Some loss models calculations, particularly numerical integration for goodness-of-fit tests or solving likelihood equations by iteration, become tedious on these calculators. Learning how to use the statistical functions and the solver properly saves time, but you will still be slower than you would be on a full graphing calculator. Practice with the approved devices early so you are not surprised on exam day. Focus on loss models first. Get comfortable with truncation and censoring adjustments. Do timed practice problems until the process feels automatic. That is the practical path through the Soa Exam Fam L portion, and it is the one that actually works.
