Errata for 2019 Exam 7 Products

High Level Summaries & Flashcards

Brehm (3/2/19) - On page 105, the formula for the Marginal Co-Measure has a typo. The numerator should have a plus sign before the epsilon and be: rho(Y + εXj) − rho(Y)

Taylor (11/17/18) - The formula for the standardized deviance formula originally showed the square root in the denominator. I corrected the formula to match what’s in Taylor. I updated the formula in the version on the CFRF online seminar and it’s corrected for all orders after 11/17.

Taylor (11/30/18) - On page 50, the formula for the mean (mu) when p = 3 is incorrect. It should be (-2*theta)^-1/2. Also, it’s the Inverse Gaussian distribution (not the Inverse Gamma).

Practice Problems

RF Taylor - 2 (4/23/19) - Just a note of clarification: Like the Cookbook recipe shows, f - 1 is the correct variable for the GLM. I plan to change that in the problem 2 solution for next year.

There's a little confusion in the paper which was why I had shown it originally as [f1, f2, …]. If you look on pg. 28, it shows that f - 1 is the variable for the model (which would go in the beta matrix). But on pg. 30 it shows the beta matrix of just [f1 , f2, ...]. I think the disagreement is simply that we're really focused on the LDF so maybe the "-1" was dropped on pg. 30 just for a cleaner look.

RF Shapland - 5b (4/17/19) - Just a note of clarification: The negative value adjustments are really only needed when we're using the GLM approach and need the log-link triangle.

For next year, I'll clarify that in this problem and re-work part b to reflect that we are making the adjustment for the incremental values that go into the GLM.

Shapland does point out on pg. 21 that when we do the adjustments for the negative incremental values that go into the GLM, the final result from the true GLM and the "simplified GLM" will be different.

RF Teng - 3 (4/9/19) - The basic premium factor should be labeled “Average Basic Premium Factor (including tax)”.

Meyers - 1 (1/2/19) - The percentile of outcome for dataset 6 in the problem should be 86.2 instead of 14.4. This matches the data used in the solution.

Siewert - 8 (1/3/19) - The losses given in the problem should be paid losses, not reported losses:

  • Unlimited paid losses for accident year 2014 are $212,000

  • Paid losses limited to $500k for accident year 2014 are $178,000

Exam 7 Cookbook

Robbin IRR paper (4/23/19) - Clarification: The loss reserve formula should really be the following:

  • Loss Reserve = sum(incurred losses to date) - sum(paid losses to date)

Both formulas result in the same loss reserve as long as 100% losses are incurred at t=1.

Patrik - IBNR Monitoring (2/2/19) - The table on Step 2 should state: “Predicted Reported Loss at March 31, 2016.