ch6 doubts

Comments

• Graham

  March 2021

  Question 1: The wiki article says CATs are high frequency low severity events (with a high total loss). But, isn't that the definition of attritional claims? On an aggregate basis, even att. claims can make a high total loss depending upon the amount of exposure to risk. Aren't CATs supposed to be low frequency high severity events?

  • The catastrophe itself, like a hurricane, is considered a low-frequency, high-severity event. But claims made against an insurer will be high-frequency because something like a hurricane could affect millions of people. The individual claim severities might vary greatly depending on whether someone's house was destroyed or whether they just had something like a broken window.

  Question 2: Fall 2014 (Q4) part b. --> We have asked the indirect effect of just the maximum benefit change in isolation on frequency and duration. Had they asked the indirect effect of the combined change (i.e max benefit level DECREASE as well as compensation rate INCREASE), would we able to tell the impact?

  • It would depend on the magnitude of the changes because the changes you described have opposite effects. If one change was large and the other small, then you still might be able to draw a valid conclusion. If the changes were roughly equal however, you probably couldn't draw a conclusion based on qualitative reasoning.

  
  
• Graham

  March 2021

  Question 3: Fall 2016 (Q4) part c. --> It states in the sample answer that due to fewer higher deductible policies being sold in PY 2014, frequency trend increases and severity trend decreases. But, how does that lead to the conclusion that PP trend and thus avg ultimate loss calculated in part a. increases? Unless of course, we assume that insurer pays smaller claims without much investigation.

  • Let's simplify the problem slightly and assume that everyone had a deductible of $500 prior to PY 2014. That means the insurer was responsible for all loss amounts in excess of $500. If everyone then switched to a $250 deductible starting with PY 2014, the insurer will still be responsible for losses in excess of $500 but now also responsible for losses in the $250 to $500 range. In other words, the insurer's TOTAL annual ultimate loss will be greater than before the change by the amounts that fall into the $250 to $500 range.
  • It's true that frequency increases and severity decreases (severity is average loss) but the net effect is that the insurer is responsible for a broader range of claims so the TOTAL loss increases.

  Question 4: Fall 2015 (Q4) --> The latest data is of Quarter 4 of CY 2014 as mentioned in the question. (Exact lines -  ) So, shouldn't the historical trend period for trending frequency end at 15.11.2014 instead of 1.7.2014?

  • I see what you mean because whoever did the trend fits must have had quarterly loss data, although no raw data was provided in the problem, quarterly or otherwise. I checked several other exam problems that dealt with 2-step LOSS trending and they all assumed the "trend-to" period was 7/1 rather than 11/15. So when applying 2-step trending to loss data, and you are not actually given the raw data, it's probably safer to assume you're dealing with years not quarters.
  • Note also that even though the trend fits were done at the end of quarters, the data was not quarterly. Each data point in the 4-point fit used 12 months of data, but it was the prior 12 months for each of 4 quarters.
  • Anyway, it would have been clearer of the question had somehow indicated whether you were dealing with quarterly or annual data but like I said above, it seems from other exam problems that you should assume annual data unless otherwise stated.

  
  
• Graham

  March 2021

  Question 5: Spring 2015 (Q8) part a. --> In Sample Answer 5, what does the following line mean - "...a short-term provision will be large in the years following many large losses and small in years following times with fewer large losses." ?

  • If you have a lot of large losses in say 2012 (2012 excess ratio = 6.3% in this problem) and base your excess ratio on just 2012, your large loss load will be too high in the subsequent years. But if you consider only 2014, for example, which has an excess ratio of only 2% then your large loss load in years subsequent to 2014 would probably be too low.
  • By the way, when the examiner's report has so many sample answers (9 for this problem) you might be slowing yourself down if you go through all of them. You will certainly learn the material a little more deeply if you do that but you then put yourself in danger of running out of study time before the exam. My recommendation is that if you read and understand 1 or 2 of the sample answers and skip the rest, that will be a more effective study strategy than trying to understand all 9 sample answers. Because there is so much to study, you have to consider the trade-off between speed and depth. You can always make a note and come back to a question later, provided you have time.

  Question 6: Fall 2013 (Q5) part a. --> Can I assume exposure base is not OLEP given but something else which may be inflation sensitive and thus multiply the freq, sev and PP trend to get the annual loss trend factor? In that case I'll leave the OLEP unadjusted in the denominator of LR and would get a different answer.

  • I'm not sure you would get full credit if you did that. It's true that the question didn't specifically state that the exposure base was earned premium, but if you assume it's something else and don't apply the exposure trend to premium, then there is no place to include it. I definitely think they were testing to see whether you knew that an inflation-sensitive exposure based needs to be trended. For that reason, your best assumption to take the given information at face value and make the assumption that EP is indeed the exposure base.

  
  
• Graham

  March 2021

  Question 7: Spring 2013 (Q7) --> The examiners' report says the following in the last paragraph - Several candidates correctly calculated the average benefit level for losses in each of the given accident years, but then multiplied the given losses by the average benefit level (rather than using the average benefit level to calculate a benefit level adjustment factor before applying) So, does that mean not accounting for the fact that losses given are all before benefit level changes and using the usual CLL/ALL approach to find CLLF was acceptable?

  • I would ignore this comment in the examiner's report. Just make sure you understand the correct solution and move on. I don't think there is much value is spending time trying to figure out the incorrect answer, especially since they didn't provide the numerical details and the problem is from way back in 2013. It's a more effective use of your time to push ahead through the material and focus on more recent exam questions that are likely more representative of what you'll see on upcoming exams.

  Question 8: Spring 2013 (Q8) --> While selecting the historical freq and sev trends, ideally, shouldn't we fit the exponential curve to points which DO NOT include the book of business changes starting 1/1/11 (i.e may be using the starting 8 points and not the latest 8 points to fit the curve and make a selection)? Isn't that technically more correct?

  • That's not correct. In general, it's better to use more recent data. Also, in this case you do want to use data after the change because that's more indicative of what's likely to happen in the future.