PDA

View Full Version : Exam 5 Pass Mark


the yellow dart
06-28-2004, 07:30 PM
Got my exam analysis back. Failed with a 5. Got a minimum of 53.75 points. No idea if this minimum would translate to a score of 5.

J.T.
06-29-2004, 10:29 AM
Got my exam analysis back. Failed with a 5. Got a minimum of 53.75 points. No idea if this minimum would translate to a score of 5.

Same here, but my minimum was a 56.375 from looking at the grade report.

Has anyone ever appealed to get a few questions regraded? If so, could you explain the process (and how long it takes). TIA!

snafu
06-29-2004, 11:13 AM
Got my exam analysis back. Failed with a 5. Got a minimum of 53.75 points. No idea if this minimum would translate to a score of 5.

Same here, but my minimum was a 56.375 from looking at the grade report.

Has anyone ever appealed to get a few questions regraded? If so, could you explain the process (and how long it takes). TIA!

I've appealed multiple questions a couple of times. Twice I have received additional points from the appeal, but alas in neither case were the points enough to change my FAIL to a PASS. Heresay has it that one of my appeals did result in someone else's status change.

The process I followed was to write a very polite email with a clear section for each question. Be clear and state your case for an alternate answer to the solutions. Check the deadline in the syllabus. It is sometime after the solutions have been published. Send your email to bcraver@casact.org. Although none of my appeals were fruitful, Mr. Craver is very helpful and will make sure that your appeal gets to the exam committee.

Then get busy studying for your next exam ;)

Good Luck

Maine-iac
06-29-2004, 11:24 AM
My appeal was not related to a specific exam question, but I will second the motion that Mr. Craver is a very polite and helpful person when you are making an exam-related petition.

Maine-iac
06-29-2004, 11:41 AM
That's the problem. You can't really write a good appeal, unless you remember the approach you took on a question.

If you see the sample solutions, and you remember your solution as taking a different approach, or including other items, but you believe it to be equally valid, then you say: "on question X, I believe my answer of Y is a valid alternative solution to the question as posed." Then explain why what you said is just as good.

It is very difficult to remember what you did for a given question. Apparently, the CAS sees a lot of appeals where what the candidate thinks they wrote is NOT what they really wrote. However, they fear that if they returned a copy of the candidate's actual answers to the candidates, they would be buried in frivolous appeals.

2bActuary
06-29-2004, 11:49 AM
But that is not a fair process. How can we expect self to remember answers to some questions 8 weeks ago? I was busy writing any answer I could come up with then, I did not know I am supposed to remember them just in case...Somebody give me a break!!!

Howard Mahler
06-29-2004, 12:17 PM
I believe actual answers are not returned because it would be another step involving extra work.

Grading essay tests already puts a very large strain on the volunteer resources. Writing questions, putting together the exam, proofing the exam, grading the exam, and deciding on the pass mark, those are the required basics.

The CAS finds time for some important extras such as posting answers to the essay questions along with the past exams. Most current students probably take this for granted, but it was a big effort to make this a regular part of the process. It is not a trivial amount of work.

Howard Mahler


Apparently, the CAS sees a lot of appeals where what the candidate thinks they wrote is NOT what they really wrote. However, they fear that if they returned a copy of the candidate's actual answers to the candidates, they would be buried in frivolous appeals. (http://actuary.ca/phpBB/viewtopic.php?p=588381#588381)

Is this a hypothesis or do we know this for a fact? The conspiracy theorist in me says that actual answers aren't returned because this would make the grading process more transparent and possible lead to higher pass ratios, which although lip service must be given to in order to maintain credibility for the designation, is not really in the best interest of anyone actually having the designation.

MountainHawk
06-29-2004, 12:42 PM
Now, I'm a faster exam taker than most, so maybe this advice isn't feasible for others, but I always try to jot down 5-10 words in the exam booklet (and a final answer for an calculation question) so I remember how I sort of did it.

GefilteFish144
06-29-2004, 01:59 PM
If you really want to appeal, the best thing would be to follow the discussion threads on the websites after the exams. Certainly there are enough people going through all sorts of petty details that you can somewhat reproduce what you did.

Unfortunately that wouldn't work for me as it would probably put me through so much anxiety that I might not live to see my exam results. Then again, it hasn't been much of an issue for me since only 20% of my failures have been 5's.

Mel-o-rama
06-30-2004, 12:05 PM
I appealed a question last year, and I thought I had a good case on one of my questions. The appeal made its way past Mr. Craver and I got the final reply of something like, "We have reviewed your appeal and we feel that no change is necessary to your score."

I felt that they just didn't want to give me anything since they would have to regrade all papers and give the same to others. Yet, I knew I had a good case. That, more than anything, really distressed me and I lost a lot of confidence in the exam process. (Even though I've just passed an exam, my level of confindence in the exam process is nowhere near what it used to be.)

Good luck in your appeals. Might I suggest possibly starting up a discussion on the forum simultaneously to your appeal. If you can get a lot of people agreeing with you, that may work little wonders. Maybe that's what I should have done last year.

GefilteFish144
06-30-2004, 12:57 PM
A former boss of mine had his exam status changed after the fact when he benefited from a point increase after another candidate appealed. It can be done.

Howard Mahler
06-30-2004, 11:23 PM
On a number of occassions during my time on the exam committee, appeals led to additional students passing.

As a percent of the total number of students on all the exams it is not large, but it is common to have one successful appeal (affecting a handful of students) per year.

Howard Mahler

P.S. There is no quota on the number of successful appeals. There can be none or more than one.

2bActuary
07-01-2004, 01:04 PM
For those who scored 5, mind sharing your minimum with rest of us? My minimum is 54.75.

Super Silver Haze
07-01-2004, 01:28 PM
55.5 minimum here.

ala
07-01-2004, 03:54 PM
Minimum: 58.375
Maximum:74.335

Student2002
07-05-2004, 05:52 PM
54.0

N.S. Sherlock
07-06-2004, 02:22 PM
Min: 58.25
Max: 65.625

I was really upset about MC #7. I sent in a letter within the 2 weeks after the exam, but they still did not accept my answer choice of E. Any body else get E? I agree that C is one possible answer based on how you interpret it. In fact, I thought of doing it that way during the test. Unfortunately, I thought their approach was too easy and not what they intended. I interpreted the premium trend of 2.9% as a total premium trend, as opposed to an average premium trend. Since it wasn't labeled and exposure trend was given, I divided the 2% exposure trend out of the 2.9% total trend to get an average per insured premium trend. The numerator did not include exposure growth since it had frequency and severity trend (i.e. it was not a total loss trend). So, that's why I took this approach.

Based on my range of possible scores, this 1.25 point swing is probably the difference in me passing or failing.

HELP!!!

Avi
07-06-2004, 02:37 PM
Min: 58.25
Max: 65.625

I was really upset about MC #7. I sent in a letter within the 2 weeks after the exam, but they still did not accept my answer choice of E. Any body else get E? I agree that C is one possible answer based on how you interpret it. In fact, I thought of doing it that way during the test. Unfortunately, I thought their approach was too easy and not what they intended. I interpreted the premium trend of 2.9% as a total premium trend, as opposed to an average premium trend. Since it wasn't labeled and exposure trend was given, I divided the 2% exposure trend out of the 2.9% total trend to get an average per insured premium trend. The numerator did not include exposure growth since it had frequency and severity trend (i.e. it was not a total loss trend). So, that's why I took this approach.

Based on my range of possible scores, this 1.25 point swing is probably the difference in me passing or failing.

HELP!!!

Why is this C and not B?

the yellow dart
07-06-2004, 06:23 PM
Min: 58.25
Max: 65.625

I was really upset about MC #7. I sent in a letter within the 2 weeks after the exam, but they still did not accept my answer choice of E. Any body else get E? I agree that C is one possible answer based on how you interpret it. In fact, I thought of doing it that way during the test. Unfortunately, I thought their approach was too easy and not what they intended. I interpreted the premium trend of 2.9% as a total premium trend, as opposed to an average premium trend. Since it wasn't labeled and exposure trend was given, I divided the 2% exposure trend out of the 2.9% total trend to get an average per insured premium trend. The numerator did not include exposure growth since it had frequency and severity trend (i.e. it was not a total loss trend). So, that's why I took this approach.

Based on my range of possible scores, this 1.25 point swing is probably the difference in me passing or failing.

HELP!!!

Why is this C and not B?


(1.06 * .99 / 1.029) ^ 2.5 = 1.050

1.050 * 35/50 = .735

I'm fairly certain there is a similiar problem in the Jones paper. I'm too lazy to look it up at the moment.

Then again, given my results, the above could be the wrong procedure and I tripped over the correct answer by pure dumb luck.

N.S. Sherlock
07-07-2004, 11:18 AM
I agree with Yellow Dart about his solution being what they were looking for.

To clarify, here is what I did that I am appealing is an acceptable solution, as well:

[1.06 * .99 / (1.029/1.020)] ^ 2.5 = 1.104

1.104 * 35/50 = .773

takeshire
07-12-2004, 12:27 PM
Anyone who got item 16 wrong or for item 17 selected statement 3 only as true, please contact me or respond. I have ways to appeal these two items that you may find appealing!

takeshire
07-12-2004, 01:29 PM
I used to work for the Educational Testing Service, and they try very hard to make certain their answers are unambiguous. They statistically verify that their exams do correlate with the ability of students to succeed in college. And they try to make certain that exams are culturally neutral.

The way the CAS exams are now administered (especially given the speeded nature of them), I don't believe they truly measure examinee understanding of the material. I believe they measures other aspects of an examinee, such as ability to work under duress, ability to express themselves in the English language (quickly), and even handwriting speed.

I believe that Exam validity and equitability are also "basics."

I understand that it would take far greater resources than the CAS has to do the kind of exam preparation and statistical analysis that they do at ETS. But doesn't that make it incumbent on the CAS to make the process an open one? Certainly, it would not take much more work for the CAS to make one extra photocopy of candidate answers and send it back in the self-addressed stamped envelope that they already provide. For the CAS to say that the reason they don't is because it will take too much work seems insincere.

My guess is that this type of openness would probably eventually lead to a higher quality exam, questions that vary more in their difficulty, and an exam that more accurately measures the understanding and preparedness of candidates (with a wider range of examinee scores ... fewer in the middle). It would probably also initially mean more pointed and successful appeals by candidates until the exam committee adjusts to writing exams that more clearly demarcate between prepared and unprepared candidates.

What are your thoughts on this?

cas_student
07-12-2004, 04:32 PM
I don't think it really matters in the end if the CAS releases more information. There will always be a way to pass the number of people the CAS wants to pass.

Just like most of the people that are taking exams, I have had to revive my squashed ego after failing, but ultimately what it boils down to is knowing the syllabus extremely well. We know the process sucks, but I think the energy should be focused on our own shortcomings rather than the exam process. The competition will always be tough no matter what the CAS decides to release. Ultimately you have to do what you can to put yourself on top.

I think we delude ourselves sometimes when we fail after studying several hundred hours. Every time I fail and restudy the material, I learn more than I did before. Once I'm at the point where everything has really sunk in, I pass. I only blame myself when I don't pass.

Marvel Mole Man
07-12-2004, 04:42 PM
I don't think it really matters in the end if the CAS releases more information. There will always be a way to pass the number of people the CAS wants to pass.

Just like most of the people that are taking exams, I have had to revive my squashed ego after failing, but ultimately what it boils down to is knowing the syllabus extremely well. We know the process sucks, but I think the energy should be focused on our own shortcomings rather than the exam process. The competition will always be tough no matter what the CAS decides to release. Ultimately you have to do what you can to put yourself on top.

I think we delude ourselves sometimes when we fail after studying several hundred hours. Every time I fail and restudy the material, I learn more than I did before. Once I'm at the point where everything has really sunk in, I pass. I only blame myself when I don't pass.

Good luck getting anyone on this board to buy that one!

I agree, though. The problem with this younger generation is that they just don't want to do the work. They want everything to be handed to them. They want it to be easy.

GefilteFish144
07-12-2004, 04:52 PM
I understand that it would take far greater resources than the CAS has to do the kind of exam preparation and statistical analysis that they do at ETS. But doesn't that make it incumbent on the CAS to make the process an open one? Certainly, it would not take much more work for the CAS to make one extra photocopy of candidate answers and send it back in the self-addressed stamped envelope that they already provide. For the CAS to say that the reason they don't is because it will take too much work seems insincere.

My guess is that this type of openness would probably eventually lead to a higher quality exam, questions that vary more in their difficulty, and an exam that more accurately measures the understanding and preparedness of candidates (with a wider range of examinee scores ... fewer in the middle). It would probably also initially mean more pointed and successful appeals by candidates until the exam committee adjusts to writing exams that more clearly demarcate between prepared and unprepared candidates.

What are your thoughts on this?

You think this is bad, try reading the posts for SOA exams. Here are some of the highlights:

1. They don't get their exam booklets mailed to them.
2. If they fail, there is no exam analysis saying which questions they got wrong.
3. The most recent copies of many of the exams are not even posted, and probably will not be posted in time for the next exam.
4. There is no guessing penalty. On the latest Exam 3, some people passed by answering roughly 15 out of the 40 questions correctly and taking wild guesses at the rest. If their letter of choice was the lucky one, they passed.

The CAS exams could certainly use some improvements, but they are certainly miles ahead of what the SOA candidates have to go through.

Maine-iac
07-12-2004, 04:57 PM
All that you say is true, but that doesn't mean that the CAS should rest by the pool until the SOA catches up in terms of reasonable exam administration practices. :)

Aim higher!

GefilteFish144
07-12-2004, 05:05 PM
All that you say is true, but that doesn't mean that the CAS should rest by the pool until the SOA catches up in terms of reasonable exam administration practices

As I said in the bottom of my message, the CAS could use improvements. My simple message was that for all the problems there are, it could be much worse. Could you imagine taking a 4-hour exam and not be able to bring food? Could you imagine having to take a 6.5-hour exam? We here at the CAS have a much better exam process, and according to the salary surveys I've read we get paid better too.

The time will come that I will eventually go back to griping about the CAS but for now I'm counting my lucky stars I didn't end up in life, health or pension.

Boxer
07-13-2004, 08:46 AM
I agree with GefilteFish - CAS is so much better than SOA. I switched over from the Pension track after getting through 3 exams (many more prior to the 2000 transition) plus a couple of EA exams.

It was very refreshing to get the performance report the first time I failed CAS Exam 5. At least I knew where I went wrong - it removed a lot of the mystery behind the grading process.

Getting to see recent exams is also a luxury. Under SOA, the most recent exams published were about 3 years old - totally useless for practice tests.

CAS trackers should count your blessings (however, I must say that I do not like the guessing penalty. Sometimes I avoid too many questions for fear of mistakes):

1) no project once you finish exams
2) no 6.5 hour exams
3) no pre-tests before going to the part 7 exam seminar
4) you get a copy of your exam booklet
5) you get a grade report when you fail
6) a pass is a pass - people that get 10s don't look down on those of us that scrape by with 6's
7) you don't have to pick a specialty for fellowship exams - makes you more portable
:guitarwo:

bermi
07-13-2004, 09:51 AM
I agree with GefilteFish - CAS is so much better than SOA. I switched over from the Pension track after getting through 3 exams (many more prior to the 2000 transition) plus a couple of EA exams.

It was very refreshing to get the performance report the first time I failed CAS Exam 5. At least I knew where I went wrong - it removed a lot of the mystery behind the grading process.

Getting to see recent exams is also a luxury. Under SOA, the most recent exams published were about 3 years old - totally useless for practice tests.

CAS trackers should count your blessings (however, I must say that I do not like the guessing penalty. Sometimes I avoid too many questions for fear of mistakes):

1) no project once you finish exams
2) no 6.5 hour exams
3) no pre-tests before going to the part 7 exam seminar
4) you get a copy of your exam booklet
5) you get a grade report when you fail
6) a pass is a pass - people that get 10s don't look down on those of us that scrape by with 6's
7) you don't have to pick a specialty for fellowship exams - makes you more portable
:guitarwo:

You sold me!

There is no doubt that improvements can be made in the current CAS structure to make the process more transparent; however, we must not forget our friends over on the SOA side who seemingly enjoy making life a living h*ll for there studets. Our system might be far from perfect, but it is 10 times better then that of the SOA!!

GefilteFish144
07-13-2004, 10:08 AM
(however, I must say that I do not like the guessing penalty. Sometimes I avoid too many questions for fear of mistakes):


My advice is don't be afraid of the guessing penalty. Think of it this way: if you get it right you get a dollar. Otherwise you lose a quarter. So if you can narrow it down to 2 possible solutions, it pays to guess. The guessing penalty is there to avoid the SOA fiasco where people passed exams by answering less than half the questions and putting a single letter down for the remaining answers.

takeshire
07-13-2004, 10:38 AM
.a

takeshire
07-13-2004, 10:41 AM
First of all,

Anyone reading the site, please don't forget that I want to team up to appeal questions 16 and 17.

Secondly,

I never advocated an easier exam. I suggested that returning exams would not be expensive or time consuming, and it would make the CAS more accountable to the candidates for fairness and exam quality.

How could a candidate not advocate fairness and exam quality? Wouldn't you rather your exam performance be a measure of your preparation than statistical noise?

With a 4-hour speeded exam, one of the main foci is test-taking stamina. Is test-taking stamina truly a relevant measure of a qualified actuary?

Item gradation (easy to extremely difficult) within a relatively unspeeded exam is one way to increase an exam's discriminating power (discriminating between well-prepared and less well-prepared candidates) while reducing the dominance of test-taking stamina as a factor in exam performance.

Father of two
07-13-2004, 10:48 AM
What is you appeal basis for #16? I missed that one.

takeshire
07-13-2004, 02:04 PM
What is you appeal basis for #16? I missed that one.

Father of two, I sent you a personal response. Check your personal mailbox.

:judge:

Avi
07-13-2004, 03:27 PM
I never advocated an easier exam. I suggested that returning exams would not be expensive or time consuming, and it would make the CAS more accountable to the candidates for fairness and exam quality.

How could a candidate not advocate fairness and exam quality? Wouldn't you rather your exam performance be a measure of your preparation than statistical noise?

It opens the can of worms of "Why is my answer worse than his answer?" I am not speaking from knowledge, but I believe that it is very possible that the graders can look at one essay and feel that the candidate has a better understanding than another - although it may be difficult to point to discrete instances and correlated point values.

With a 4-hour speeded exam, one of the main foci is test-taking stamina. Is test-taking stamina truly a relevant measure of a qualified actuary?

You've never been through a reinsurance 1/1 renewal season, have you. Once you have, you will see how performance under pressure is actually a job reqirement for an actuary. And while that may not be within the purview of the CAS to test, it DOES have a correlation with the ability of the candidate to function as an actuary.

ahow
07-13-2004, 03:28 PM
I still haven't gotten my dang score in the mail yet!!! :swear: I just called the CAS yesterday and had them resend it though...

anaglyph
07-13-2004, 03:32 PM
I answered C for #17. I got a 5 on the exam. I spent hundreds of hours studying. It is certainly worth my time to apeal what ever I can. What is your argument for why the answer to #17 is C?

takeshire
07-13-2004, 04:17 PM
It opens the can of worms of "Why is my answer worse than his answer?" I am not speaking from knowledge, but I believe that it is very possible that the graders can look at one essay and feel that the candidate has a better understanding than another - although it may be difficult to point to discrete instances and correlated point values.

You've never been through a reinsurance 1/1 renewal season, have you. Once you have, you will see how performance under pressure is actually a job reqirement for an actuary. And while that may not be within the purview of the CAS to test, it DOES have a correlation with the ability of the candidate to function as an actuary.



In response to your first comment:
I absolutely agree that graders could do that. But that's quite subjective, isn't it? Remember, this is business mathematics that we're talking about here, not creative writing.

Regarding that part of your statement, I would ask the following questions:

"How much human error / oversight occurs in the grading process?"

"How much error do you think is allowable, given the time and money invested by each candidate?"

"How would you propose to meet that standard?"


In answer to your second comment:
No, I have never been through a reinsurance 1/1 renewal season, but I have worked for a consulting firm and a reinsurance company, and I personally know several fine actuaries who perform well at the job, and who know the exam material quite well, but who still repeatedly fail exams (and not for lack of knowledge of the material). Perhaps you don't.

takeshire
07-13-2004, 04:20 PM
I answered C for #17. I got a 5 on the exam. I spent hundreds of hours studying. It is certainly worth my time to apeal what ever I can. What is your argument for why the answer to #17 is C?

I'll send you a personal response. Please be on the lookout for it.

Boxer
07-13-2004, 04:57 PM
I also put C for #17 and wanted to complain, but I found the correct answer straight out of the text in Tiller.

Check out Tiller and see if you think an appeal would work - I don't think so.
:viola:

GA Peach
07-14-2004, 09:42 AM
True Boxer, but remember, just because it's in the article doesn't mean it's the gospel.

takeshire
07-14-2004, 10:37 AM
True Boxer, but remember, just because it's in the article doesn't mean it's the gospel.


I didn't want to post this because it is so long, but since so few people are responding to my private messages, below is my appeal of item 17. If anyone can refute this argument on its own grounds, I would appreciate a response. Otherwise, good luck with your appeals:


For item 17, the question is worded as follows:

Which of the following statements are true concerning ISO’s Commercial General Liability (CGL) Experience Rating Plan?

1. The maximum single loss (MSL) limitation makes the plan more responsive to severity than frequency.
2. The MSL increases as the size of the risk increases.
3. The D-Ratio adjusts for the impact of the MSL.


I immediately thought that statement 2 is not always true. The formula for the MSL, according to Tiller Sherwood, is as follows:

MSL = 0.3 * E / Z, where E = the expected losses, and Z = the credibility factor

Considering a set of 20 risks, since the MSL depends on both the expected losses and the credibility factor, it is clear that there are cases where the size of the risk can increase, while the MSL decreases.

The following two statements would have always been true:

The MSL tends to increase as the size of the risk increases.
For an individual risk, the MSL increases as the size of the risk increases.

The above is a complete argument, and was my line of reasoning on the exam, but the argument below may further clarify my reasoning.

As quoted from Tiller:

“For the plan as revised in July 1988, the MSL is set so that the maximum impact of any single occurrence is +0.30 on the resulting experience modification factor. Consequently, MSL’s increase as the size of the risk (as measured by premium or credibility) increases.”

Embedded within these two sentences is an implicit assumption that we are speaking about an individual risk. It is always true that the MSL increases as the size of the risk increases only if it is understood that you are talking about an individual risk.

Taken alone, statement 2, “The MSL increases as the size of the risk increases,” does not make it clear that we are talking about only one risk. There are two possible ways of interpreting this sentence.
A. Assume we are speaking of a single risk.
B. Make no assumption, in which case we may be speaking of a single risk or of a number of risks.
In case B, statement 2 implies that whether the MSL increases or decreases is dependent solely on the size of the risk, which is false.

Considering a set of risks ordered from lowest to highest on a premium basis, it is possible that the MSL does not increase as the size of the risk increases, since the MSL is based on the credibility factor, which varies from risk to risk. A mathematical relationship is said to be true if and only if it is always true. Therefore, given the “no assumption” interpretation (option B, above), C is the correct response.

Colymbosathon ecplecticos
07-14-2004, 10:51 AM
I would disallow this appeal.

Statements such as your #2 are meant to be taken ceteris paribus, meaning that other variable are held constant. In your example, the other risks would need to have comparable credibilities and expected losses --- if they do, the statement is true.

Me3
07-14-2004, 12:48 PM
Taken alone, statement 2, “The MSL increases as the size of the risk increases,” does not make it clear that we are talking about only one risk. There are two possible ways of interpreting this sentence.
A. Assume we are speaking of a single risk.
B. Make no assumption, in which case we may be speaking of a single risk or of a number of risks.
In case B, statement 2 implies that whether the MSL increases or decreases is dependent solely on the size of the risk, which is false.


I'm not sure what about the statement allows you to interpret it as possibly referring to more than one risk.

It says:
"the size of the risk increases" = one risk ("risk" & "increases" are both singular terms)

It does not say:
"the size of the risks increase" = more thank one risk

takeshire
07-14-2004, 02:25 PM
I would disallow this appeal.

Statements such as your #2 are meant to be taken ceteris paribus, meaning that other variable are held constant. In your example, the other risks would need to have comparable credibilities and expected losses --- if they do, the statement is true.


Thanks for your response:

If what you said was the intent of the item writer on this question, the exam committee should be willing to throw the entire question out. Clearly, when the E variable (which represents premium, or exposure, or expected losses) moves, the Z variable (credibility) is also going to move. Tiller's comment was in reference to her definition of credibility which disallowed Z to move more quickly than E, but only for a single risk (which is the only way that Z can be defined).

Any candidate who truly understood Tiller's chapter and study note, would be aware that the statement was referring to a single risk (and that since the variables both are moving, the ceteris paribus assumption is not relevant). As the single risk aspect of the statement was not spelled out on the exam question, it is clearly an incomplete, thus a false statement.

takeshire
07-14-2004, 02:37 PM
It says:
"the size of the risk increases" = one risk ("risk" & "increases" are both singular terms)

It does not say:
"the size of the risks increase" = more thank one risk[/quote]


Thanks for your response:

Consider the following two sentences:
Given a set of 20 risks, the MSL increases as the size of the risk increases.
Given a single risk, the MSL increases as the size of the risk increases.

Colymbosathon ecplecticos
07-14-2004, 02:55 PM
I won't debate this with you. In my view this is an example of a frivolous appeal. It will take some of the committee's time that could be spent on a valid appeal and waste it instead.

Suppose that you succeed in getting the question thrown out. Some candidates that now have very low 6's may now have very high 5's --- they might not be very happy with the results of your frivolous appeal.

takeshire
07-14-2004, 03:06 PM
I won't debate this with you. In my view this is an example of a frivolous appeal. It will take some of the committee's time that could be spent on a valid appeal and waste it instead.

Suppose that you succeed in getting the question thrown out. Some candidates that now have very low 6's may now have very high 5's --- they might not be very happy with the results of your frivolous appeal.


I would only agree that my appeal is frivolous if my interpretation of the question was not valid. If the question is thrown out, or if an alternate answer is given, then clearly the committee will not have considered my appeal frivolous.

As far as candidate's scores being lowered, it is my understanding that once you've passed, you've passed. No one is going to change your pass to a fail. Appeals can only result in additional candidates passing. Not the reverse. I have called the CAS and this is confirmed. Don't worry about your low six. It's safe. :shake:

GefilteFish144
07-14-2004, 03:06 PM
Suppose that you succeed in getting the question thrown out. Some candidates that now have very low 6's may now have very high 5's --- they might not be very happy with the results of your frivolous appeal.

Has anyone ever heard of this happening? I personally know someone who passed an exam after a successful appeal, but I've never heard of anyone "un-passing" an exam. I don't think even the SOA would be that cruel, much less the CAS.

ramanujan
07-14-2004, 03:53 PM
True Boxer, but remember, just because it's in the article doesn't mean it's the gospel.


I didn't want to post this because it is so long, but since so few people are responding to my private messages, below is my appeal of item 17. If anyone can refute this argument on its own grounds, I would appreciate a response. Otherwise, good luck with your appeals:


For item 17, the question is worded as follows:

Which of the following statements are true concerning ISO’s Commercial General Liability (CGL) Experience Rating Plan?

1. The maximum single loss (MSL) limitation makes the plan more responsive to severity than frequency.
2. The MSL increases as the size of the risk increases.
3. The D-Ratio adjusts for the impact of the MSL.


I immediately thought that statement 2 is not always true. The formula for the MSL, according to Tiller Sherwood, is as follows:

MSL = 0.3 * E / Z, where E = the expected losses, and Z = the credibility factor

Considering a set of 20 risks, since the MSL depends on both the expected losses and the credibility factor, it is clear that there are cases where the size of the risk can increase, while the MSL decreases.

MSL = 0.3*E(Loss cost)*EER/Z

Z = E(loss Cost)/[E(Loss Cost) + K]

=> MSL = 0.3*[E(loss cost) + K]*EER

K is a constant. EER depends on the d-ratio curve, increases with MSL.

=> MSL increases with E(Loss Cost) which I interpret as the size of the risk. Hence Proved.

This is true even if you are comparing two different risks.

takeshire
07-14-2004, 05:59 PM
Ramanujan,


I got my formula as follows:

From page 2 of Tiller, "the MSL is set so that the maximum impact of any single occurrence is + 0.30 on the resulting experience modification factor."

from page 169 of Tiller Sherwood,

M = [(AxZ) + (E x (1-Z))]/E

This implies that

0.3 = MSL x Z / E, i.e. MSL = 0.3 E / Z.

From Tiller, page 3, Z = P / (P+K), where P is subject basic limits earned premium, but Tiller Sherwood on page page 163 also makes reference to chapter 8 as a source of credibility information, and we all know about bayesian credibility. But continuing strictly with Tiller's words:

MSL = 0.3 E (P+K) / P.

Inasmuch as two risks with the same expected losses can have different premiums, doesn't that confirm that sometimes MSL can decrease as the size of the risk increases? Tell me what I'm missing, please.

Also, where did you get the formula:

MSL = 0.3*E(Loss cost)*EER/Z

I'm not sure where the E(Loss cost) came from.

Thanks.

ramanujan
07-14-2004, 06:53 PM
Ramanujan,


I got my formula as follows:

From page 2 of Tiller, "the MSL is set so that the maximum impact of any single occurrence is + 0.30 on the resulting experience modification factor."

from page 169 of Tiller Sherwood,

M = [(AxZ) + (E x (1-Z))]/E

This implies that

0.3 = MSL x Z / E, i.e. MSL = 0.3 E / Z.

I agree with you till here.


From Tiller, page 3, Z = P / (P+K), where P is subject basic limits earned premium, but Tiller Sherwood on page page 163 also makes reference to chapter 8 as a source of credibility information, and we all know about bayesian credibility.

In this formula, E(Loss cost) = Premium*Basic Limits ELR
Adjust K with the same factor to get Z = E(loss Cost)/[E(Loss Cost) + K']

Inasmuch as two risks with the same expected losses can have different premiums, doesn't that confirm that sometimes MSL can decrease as the size of the risk increases? Tell me what I'm missing, please.

By saying that "two risks with the same expected losses can have different premiums", you are assuming that the basic limits ELR is different for two different risks. ELR = 1 - expense ratio, does not generally vary by company (at least in the same state).

If you are ready to assume things like this, there is no statement that is absolutely true. You can contest every statment on this exam.

Also, where did you get the formula:

MSL = 0.3*E(Loss cost)*EER/Z

I'm not sure where the E(Loss cost) came from.

Thanks.

E in your formula is E(basic limit losses limited by MSL) which is equal to E(basic limit loss cost)*EER.

mathseal
07-14-2004, 06:56 PM
Takeshire,
You're dead-set on this appeal, so go ahead and try it.

I think Coly's messing with your head - to the best of my knowledge no one has ever "unpassed" or even had a score reduced. Years ago I was amazed that 6 Months after an exam the CAS declared a question invalid and retroactively made a whole bunch of people associates. But short of finding fraud, I don't think the CAS wouldn't invalidate an exam credit it already released.

But I agree with Coly (this could be a first?) that your wasting their time. While I sympathize that you misunderstood the exam's intent, you're not getting a good reception from us - and we're a far more sympathetic audience than the CAS will be.

If I was on the committee I'd be asking why is this person arguing for a twisted interpretation of multiple choice question now, rather than during the "defective question" stage when the exam key came out.

The world could do without actuaries who respond to their bosses:
"Wait before you go and write that account, its not quite true that writing more profitable business would help our bottom line. I can envision a case where writing more profitable business would actually work against us by inviting greater competition."
True? yes. But more often than not useless.

takeshire
07-14-2004, 08:50 PM
Ramanujan,

You almost have me convinced. I guess the underlying thing that bothers me about this is that it means that regardless of how uncertain a book of business is, it will still receive the same MSL.

Why is a risk loading not part of the premium formula?


Thanks,
Takeshire

ramanujan
07-14-2004, 09:47 PM
Ramanujan,

You almost have me convinced. I guess the underlying thing that bothers me about this is that it means that regardless of how uncertain a book of business is, it will still receive the same MSL.

Why is a risk loading not part of the premium formula?


Thanks,
Takeshire

In context of experience rating, there is no book of business. The experience mod is calculated for a single account and applied to the manual premium.

For risk loading, I do not know what risk you are talking about. If you mean higher average cost, then that will be reflected in a higher manual loss cost. If you are concerned about variability in the loss cost for a particular class code, that is an issue for rate bureaus to tackle.

This is my understanding, but I am no expert on the subject with minimal experience.

takeshire
07-15-2004, 10:24 AM
Well, since I got some constructive feedback (or destructive, I'm not sure yet) here's my item 16 appeal. Tell me what you think. Do I have a case? Ramanujan? Coly? Bueller? Anyone?


For item 16, the question is worded as follows:

Based on Krakowski, “Quantifying the Impact of Non-Modeled Catastrophes on Homeowners Experience,” which of the following are problems with most current non-modeled catastrophe ratemaking methodologies?

1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes.
2. Only non-wind catastrophes such as fire, explosion, and water are considered.
3. No consideration is given to changes in exposure concentration over time.


The wording of item 16 makes it appear that statements 1, 2 and 3 would relate directly to the catastrophe ratemaking methodology. In that case, the following sentences illustrate an interpretation of statements 1, 2, and 3:

1. With most current non-modeled catastrophe ratemaking methodologies, trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes.
2. With most current non-modeled catastrophe ratemaking methodologies, only non-wind catastrophes such as fire, explosion, and water are considered.
3. With most current non-modeled catastrophe ratemaking methodologies, no consideration is given to changes in exposure concentration over time.

For statements 2 and 3, this is clearly the intended interpretation, thus it is reasonable to assume that this is also the intended interpretation for statement 1.

Krakowski notes the following:

"The non-catastrophe losses that form the base for the excess ratio comprehend multiple perils. Trends in some of these perils, such as liability and crime, may have no correlation to catastrophes, and cause distortion in excess ratios. Thus if liability losses become a much greater proportion of all coverages, then the ratio of catastrophe to non-catastrophe losses will artificially appear to go down, all other things being equal."

Krakowski’s statement implies that trends in losses for non-catastrophe perils, are being used in a way that assume some correlation with catastrophes.

This confirms that statement 1 is also a problem with most current non-modeled catastrophe ratemaking methodologies. Accepting this, for item 16, none of the choices, A through E are correct, since both statements 1 and 3 are problems with most current non-modeled catastrophe ratemaking methodologies.

Colymbosathon ecplecticos
07-15-2004, 10:42 AM
My first reaction was "you must be kidding", but I will answer this as though it were a real question.

Look at item 1. Is it true that the correlation is a problem? The author states that the lack of correlation might be a problem.

ramanujan
07-15-2004, 12:37 PM
My first reaction was "you must be kidding", but I will answer this as though it were a real question.

Look at item 1. Is it true that the correlation is a problem? The author states that the lack of correlation might be a problem.

I agree.

N.S. Sherlock
07-15-2004, 02:32 PM
For #16, I read statement 1 quickly and misinterpreted it. I doubt this misinterpretation would result in a successful appeal, but I read correlated as the non-statistical meaning ("to present or set forth so as to show relationship") and that they were implying that the data was just problematically being lumped together. Again, without the time pressure I think the statement is not one of the problems. But, that just goes back to the discussion about the problem of time pressure on these exams. Doh!

By the way, no one ever addressed my appeal for item 7 on page 1 of this thread. Can anyone shoot it down? Anyone agree with me?

Takeshire, sorry about not respoding before. I don't follow your argument for #17.

Colymbosathon ecplecticos
07-15-2004, 02:38 PM
NSS, you don't work in pricing, do you.

The premium trend is the increase per unit of exposure. This is simply how English works. When I say that the cost of gassoline went up 10% last year, I am not saying that the total dollar sales of gasoline went up 10%.

Wigmeister General
07-15-2004, 02:40 PM
I'm glad I don't take these things anymore. I'm really glad.

MathGuy
07-16-2004, 08:22 AM
Here's the strategy I've used on exams five and six: I skip all the multiple choice questions, then do all the essay-types. I then go back and answer a handful of MCs that I am sure of (unless I'm desperate, I've decided). That way, when results come out, I don't P&M that I got something wrong because the problem was ambiguous (even though the only reason it was ambiguous was because I didn't understand the problem).

takeshire
07-16-2004, 10:11 AM
Scenario

A bucket has 30 red balls, 30 green balls, and 10 white balls.

A computer program is designed to simulate a random draw of 14 balls from this bucket. However, the program always picks the same number, N red balls and green balls, and 14 - 2N white balls.


Which of the following is a problem with the computer program?

1. There are always the same number of red balls as green balls.
2. The number of red balls and green balls is not always the same.


Come on, guys. Open your eyes. This is an English language problem. Not a problem of knowledge of the material. Coly, read the question. Do you really believe this is a true/false question?


Coly, according to your line of reasoning, since according to the scenario the draw is random, number 1 isn't true, so it's not a problem with the computer program. You gave Sherlock a hard time about the English language. You're not perfect yourself, buddy.

Colymbosathon ecplecticos
07-16-2004, 10:23 AM
"1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."

Please explain why you believe that this is true.

My best attempts:

"Gee, there is a hurricane coming, I think I'll sue somebody"

"I think that there might be an earthquake next year, I'd better steal something today".

Basso
07-16-2004, 10:33 AM
"1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."

Please explain why you believe that this is true.

My best attempts:

"Gee, there is a hurricane coming, I think I'll sue somebody"

"I think that there might be an earthquake next year, I'd better steal something today".

liability: hurricane went through, I went to your house/store and there was debris all over the place, I fell and hurt myself! You bastage! I'm going to sue you!

Crime: tornado went through and all the police are out trying to help victims. Perfect time to go and pick up a new tv from the local store...sans payment.


I'm not saying I think they are correlated, but your best efforts are weak.

takeshire
07-16-2004, 11:00 AM
[quote="Colymbosathon ecplecticos"]"1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."

Please explain why you believe that this is true.


It's not. It's false. But as I said, it's not a true/false question.

But it is true that it is a problem with most catastrophe ratemaking methodologies. Within them trends in losses for non-catastrophe perils, such as liability and crime, are correlated to castastrophes. Just like within the computer program, there are always the same number of green and red balls.

But explaining the English language over and over is becoming a waste of time. I'm outta here.

MathGuy
07-16-2004, 12:00 PM
You know, if you spent as much time studying for the exam as you do arguing semantics, maybe you would have got some other problem on the exam right and passed.

takeshire
07-16-2004, 03:21 PM
You know, if you spent as much time studying for the exam as you do arguing semantics, maybe you would have got some other problem on the exam right and passed.


The reason I am spending so much time arguing semantics is to identify weaknesses in my appeal. And you have no idea how much time I spent studying.

MathGuy
07-16-2004, 03:40 PM
Not enough, apparently.

If A is false, then A cannot be a "problem" with anything, because it doesn't exist.

Which of these is a problem with wild game hunting?

I. Unicorns are very fast, making them a tough target.
II. ...
III. ...

Sounds like you would argue that I is true because, if unicorns existed, they would be a tough target.

takeshire
07-16-2004, 04:31 PM
Not enough, apparently.

If A is false, then A cannot be a "problem" with anything, because it doesn't exist.

Which of these is a problem with wild game hunting?

I. Unicorns are very fast, making them a tough target.
II. ...
III. ...

Sounds like you would argue that I is true because, if unicorns existed, they would be a tough target.


By that argument, since it is not true that a random draw always has the same number of red balls and green balls, the fact that the computer program picks the same number of red and green balls is not a problem.

The other reason that I continue to respond to these comments, is to demonstrate to other candidates who got the item wrong that there is a reasonable appeal to this item. My hope is that if a sound argument is published (and well-defended) on this web-site, more candidates will be comfortable appealing an item, and the CAS exam committee will take a harder look at those appeals.

I also hope that other candidates will strengthen their appeals by seeing the type of nay-saying thinking that exists out there. After all, these responses to what I have published are coming from candidates who eventually may be on the CAS exam committee (scary thought). I hope that by then they will be a little more fair-minded.

MathGuy
07-16-2004, 04:58 PM
I think you are confusing yourself, so I won't pour oil on the fire, if you know what I mean*. If you listed every problem with most current non-modeled catastrophe ratemaking procedures, the statement "Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes" would not be on the list. Therefore, the question is unappealable. The fact that the statement is false just means that it is not a problem with any thing in the world.











* You may not know what I mean, if this thread is any indication: I'm using the word "fire" as a metaphor for your problem. Throwing "oil" on your "fire" would only make it worse. If I could help you, you could say I was "throwing water on your fire."

takeshire
07-16-2004, 05:24 PM
If you listed every problem with most current non-modeled catastrophe ratemaking procedures, the statement "Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes"


It's true that in any case the question and the item are poorly worded.

If were to list every problem, this is how I would state the item in question.

"Losses for non-catastrophe perils, such as liability and crime, are trended together with catastrophes."

The statement:

"Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes"

says pretty much the same thing, except not very eloquently.

But seeing that the syntax of the question is not even correct, I think it would take a pretty closed-minded judge to not accept this as a valid interpretation.

Basso
07-16-2004, 05:55 PM
It's true that in any case the question and the item are poorly worded.
I'll give you this

If were to list every problem, this is how I would state the item in question.

"Losses for non-catastrophe perils, such as liability and crime, are trended together with catastrophes."

The statement:

"Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes"

says pretty much the same thing, except not very eloquently.
I don't think they say the same thing. The first one is stating that the inclusion of non-cat perils in the computation of a cat factor is a problem.

The second statement is asking, is the correlation between non-cat perils and cat perils a problem for the computation of a cat factor.

But seeing that the syntax of the question is not even correct, I think it would take a pretty closed-minded judge to not accept this as a valid interpretation.

Look, normally I'm much more sympathetic to the case of the person trying to make the appeal than the person arguing against it, but I just don't think you have a case. Take a look back at old CAS exams and you will see that when you have a list of items I, II, & III. If any of the items I, II or III is false, it is not correct to include it in the answer. The opposite is true if the question asks which are not, or which is false. In this case, the question is asking which is a problem. Item 1 states that there is a correllation between crime & cats. Well, there is no correlation between crime & cats, so that cannot be the problem. If the question had been worded as you propose then yes, that would be a problem, but that isn't what the question asked.

I think that if even I think your appeal is a stretch, there isn't much point in filing it, since in 99% of the cases I tend to side with the student. I'm pretty sure that the exam committees are probably biased 90+% against the students. I'm not liking your odds.

Good luck if you do file it though. (I mean this, its not a wisecrack).

takeshire
07-17-2004, 02:11 PM
New improved version of the item 16 appeal:


There are two parts to this appeal. In Part 1, I will show that grammatically, statement one should be understood as an aspect of current non-modeled catastrophe ratemaking methodologies.
In Part 2, I will show how within current non-modeled catastrophe ratemaking methodologies, statement one applies.


PART ONE:
Here, I will show that statement one should be understood as an aspect of current non-modeled catastrophe ratemaking methodologies.
Consider the following scenario:

A bucket has 30 red balls, 30 green balls, and 10 white balls. A computer program is designed to simulate a random draw of 14 balls from this bucket. However, the program always picks the same number, N red balls and green balls, and 14 - 2N white balls.

Which of the following is a problem with the computer program?

1. There are always the same number of red balls as green balls.
2. The number of red balls and green balls is not always the same.


By the wording of the question, it is clear that statement 1 is the correct statement. In other words, the interpretation that means the following:

"In the computer simulation, there are always the same number of red balls as green balls."

is the correct interpretation.


So the statement:

1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes.

should be interpreted as follows:

"In most current non-modeled catastrophe ratemaking methodologies, trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."


PART TWO:
Here I will show how non-catastrophe perils are correlated to catastrophes:

According to Krakowski,
"The ISO excess wind methodology takes as excess losses all wind losses in excess of the long term historical median ratio of wind to non-wind losses, but only for years in which that wind/non-wind ratio is in excess of 1.5 times the median ratio. These excess losses are then spread to all years."

In the above example, wind losses are catastrophe losses, and non-wind losses are non-catastrophe losses.

The maximum amount of catastrophe losses left in an accident year prior to spreading the excess losses, is thus proportional to the non-catastrophe losses.

In this manner, trends in non-catastrophe perils are correlated to catastrophes in most current non-modeled catastrophe ratemaking methodologies.


Thus both statements 1 and 3 are correct, and the item should be thrown out.

takeshire
07-17-2004, 03:20 PM
DogBoy,

I am not restating the wording of question #16. I am only clarifying what the wording means. When I said that the question is poorly worded, I was speaking softly. If I were to not pull any punches, I would say that finding statement #1 to not be a problem is 100% wrong, whatever the intended interpretation of the item writer might have been.

You said that because there is no actual correlation between crime & cats, the intentional creation of a correlation between crime and cats within the methodology is not a problem. That reasoning is clearly flawed.

-Takeshire


PS to all: There are many instances of multiple correct answers or invalid multiple-choice questions on past exams. We might expect to see a higher percentage of similarly flawed essay items, since they are more involved. I'm not sure if we do or not.

MathGuy
07-19-2004, 08:18 AM
The maximum amount of catastrophe losses left in an accident year prior to spreading the excess losses, is thus proportional to the non-catastrophe losses.

In this manner, trends in non-catastrophe perils are correlated to catastrophes in most current non-modeled catastrophe ratemaking methodologies.

Hey, I apologize for giving you a hard time on Friday. I was feeling a bit cheeky.

Two comments on your appeal (specifically the two sentensec above):

1) "The maximum amount of catastrophe losses" is a single number, so it doesn't really make sense to say that it's proportional to something. It's proportional to lots of things. What proportion are we talking about.

2) Even if you show that cat and non-cat losses are correlated, you seem to suddenly make the jump that the trends are correlated. I do not think that this is a self-evident claim, and should be supported.

just_josh-ing
07-19-2004, 10:17 AM
Thus both statements 1 and 3 are correct, and the item should be thrown out.


You've convinced me. I totally agree with you. Math Guy, I think when takeshire talks about the maximum amount of losses, he's referring to specific accident years. In that he's correct. Takeshire, you might want to also clarify that you are talking about non-catastrophe losses by accident year. And that the ratio is catastrophe losses over non-cat's (though that's clearly what we all read, it can't hurt to restate it).

Math Guy, don't correlated losses imply correlated trends? After all, if non-cat losses go up, trend is positive, and cat losses go up creating positive trend. If non-cat losses go down, trend is negative, and cat losses go down, implying negative trend. That looks like correlated to me. 8-)


JJ

Basso
07-19-2004, 10:41 AM
I am not restating the wording of question #16. I am only clarifying what the wording means. When I said that the question is poorly worded, I was speaking softly. If I were to not pull any punches, I would say that finding statement #1 to not be a problem is 100% wrong, whatever the intended interpretation of the item writer might have been.

You said that because there is no actual correlation between crime & cats, the intentional creation of a correlation between crime and cats within the methodology is not a problem. That reasoning is clearly flawed.


I think we are going to have to agree to disagree. I think you are rewording the problem to get to your conclusion. Also, I think the statement (statement 1) is clearly incorrect. To me, the unicorn example someone gave is similar to what the question here is asking. I don't think you'll convince me you are correct in the way you answered the question, but I agree whole-heartedly that the wording on these exams and the way they ask some of these questions is downright poor. Perhaps if you (and others) continue to appeal on this type of basis, the wording will get better.

Good luck.

takeshire
07-19-2004, 11:16 AM
I think we are going to have to agree to disagree. I think you are rewording the problem to get to your conclusion. Also, I think the statement (statement 1) is clearly incorrect. To me, the unicorn example someone gave is similar to what the question here is asking. I don't think you'll convince me you are correct in the way you answered the question, but I agree whole-heartedly that the wording on these exams and the way they ask some of these questions is downright poor. Perhaps if you (and others) continue to appeal on this type of basis, the wording will get better.

Good luck.


MathGuy was the one who gave the unicorn example, and he seems to be changing his mind. Perhaps you should ask him if he still thinks the unicorn argument is relevent.

By the way, did you at least read through my new improved version of the appeal? I don't see how you could have and still made the above comments.

Basso
07-19-2004, 01:17 PM
PART ONE:
So the statement:

1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes.

should be interpreted as follows:

"In most current non-modeled catastrophe ratemaking methodologies, trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."


This is the part I disagree with. I have gone from not understanding your point to seeing it somewhat, but I still disagree that it should be interpreted that way. Again, at least with this kind of focus on the wording chosen on the exams, it should improve the focus on making clear questions. I've gone from thinking it would be a waste of time for you to file the appeal to at least pointing out a problem with the exam's wording. You might even be successful. Good luck with your appeal.

takeshire
07-19-2004, 10:21 PM
Final version of the item 16 appeal:


There are two parts to this appeal. In Part 1, I will show that grammatically, statement 1 should be understood as an aspect of current non-modeled catastrophe ratemaking methodologies.
In Part 2, I will show why statement 1 is true, and that it is a problem.


PART ONE:
Here, I will show that statement 1 should be understood as an aspect of current non-modeled catastrophe ratemaking methodologies.

Consider the following scenario:

A bucket has 30 red balls, 30 green balls, and 10 white balls. A computer program is designed to simulate a random draw of 14 balls from this bucket. However, the simulation always picks the same number, N red balls and green balls, and 14 - 2N white balls.

Which of the following is a problem with the computer simulation?

1. There are always the same number of red balls as green balls.
2. The number of red balls and green balls is not always the same.

In the context of the question, statement 1 is the correct answer. In other words, the interpretation that means the following:

"In the computer simulation, there are always the same number of red balls as green balls."

is the correct interpretation. The statement is true, and it happens to also be problem.


Conversely, the interpretation that means the following:

"In a random draw, there are always the same number of red balls as green balls."

is a structurally incorrect interpretation. Though this statement is false, the falseness of the statement does not mean that statement 1 is not a problem.


The structure of the question on the exam is identical to the structure of the question above with the following substitutions:

"Historically"
replaces
"In a random draw."

"In most current non-modeled catastrophe ratemaking methodologies"
replaces
"In the computer simulation.""

"Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes"
replaces
"There are always the same number of red balls as green balls."


Thus it is clear that

1. Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes.

should be interpreted as follows:

"In most current non-modeled catastrophe ratemaking methodologies, trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."

and not

"Historically, trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes."

which though clearly false, is an imprecise (and structurally incorrect)interpretation of what is written.


PART TWO:
Here I will show how within the context of the question, trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes, and how this is a problem:

According to Krakowski,
"The ISO excess wind methodology takes as excess losses all wind losses in excess of the long term historical median ratio of wind to non-wind losses, but only for years in which that wind/non-wind ratio is in excess of 1.5 times the median ratio. These excess losses are then spread to all years."

In the above example, wind losses are catastrophe losses, and non-wind losses are non-catastrophe losses. The maximum amount of catastrophe losses left in an accident year prior to spreading the excess losses, is thus proportional to the non-catastrophe losses within that accident year.

In this manner, non-catastrophe perils are correlated to catastrophes in most current non-modeled catastrophe ratemaking methodologies. Thus the statement,

"Trends in losses for non-catastrophe perils, such as liability and crime, are correlated to catastrophes,"

is true within the context of the question.

Finally, since according to Krakowski, "Trends in some of these perils, such as liability and crime may have no [historical] correlation to catastrophes, and cause distortion in excess ratios," the truth of statement 1 is also a problem.


Thus both statements 1 and 3 are correct, and the item should be thrown out.

takeshire
07-19-2004, 10:40 PM
The above appeal is a final version that I would like to send in. It certainly is difficult and time-consuming to write a convincing appeal. Thank you everyone for all constructive criticisms, they were truly helpful, well-intentioned or not. Good luck to everyone else sending in appeals.

Hopefully the above sample appeal will help me as well as others to receive their well-earned and deserved pass marks.

My philosophy: If you believe your performance was better than it appears on your scorecard, it very well might have been. Extrapolating the past percentages of multiple answer or invalid multiple choice questions to essay questions, it may sometimes be possible that a candidate's score should have been two marks higher than what that candidate actually received. We should all be working towards a more open appeals process in which we receive all our answers so that we can defend our investments of time and emotional energy.


Takeshire

ahow
07-20-2004, 12:05 PM
I still haven't gotten my dang score in the mail yet!!! :swear: I just called the CAS yesterday and had them resend it though...FINALLY got my score. 38.5 minimum for a 2. This was my first essay exam, so it was a good learning experience...

Good luck with your appeals...

takeshire
07-20-2004, 06:22 PM
Dear Math Guy and Ramanujan:

Okay, you two, help me out here. Is my reasoning correct?


Consider risks, A and B, for the Super Efficient Company with zero expenses and ALAE* (Zero expenses and ALAE are assumed to simplify the analysis, however the results apply more broadly, ... all profit is made on investment income).

Risk A has total premium and total Expected losses of 200,000, and basic limits premium (P) and basic limits expected losses (E) of 50,000.

Risk B has total premium and total expected losses of 100,000, and basic limits premium (P) and basic limits expected losses (E) of 100,000.

Consider the formula for the MSL

MSL = 0.3 * E / Z, where Z = P / (P + 100,000)

=> MSL = 0.3 * E (P + 100,000) / P

For Risk A, MSL = 0.3 * 50,000 (50,000 + 100,000) / 50,000 = 45,000
For Risk B, MSL = 0.3 * 100,000 ( 100,000 + 100,000) / 100,000 = 60,000

Clearly, Risk A is larger than Risk B, but since the basic limits portion of the risk is smaller, Risk A has a smaller MSL.

Therefore, Tiller should have said:
"The MSL increases as the size of the risk (on a basic limits basis) increases."


Are you in agreement with this, or have I missed something in my interpretation of Tiller?

mathseal
07-20-2004, 07:33 PM
My philosophy: If you believe your performance was better than it appears on your scorecard, it very well might have been.

Takeshire,
Compared to what? If you argue that most questions are ambiguous - then everyone should score 90+ every time - but how does that help?
Is the fair solution to pass the 4's and 5's. (Which I'd guess would give a pass ratio in the 50-60%)? That's not the CAS's intent, they really want you to know this stuff cold. PROVE you are qualified (or more qualified than the rest), not that the CAS is a bad test writer. We already know that! :P


On the recent part 7, many people couldn't begin to answer a question that seemed garbled. Does it follow the question should be thrown out? I don't think so, those who somehow muddled through got some partial credit (they also spent some time that could have been used to score points elsewhere)

On the same exam there was also a multiple choice were the CAS meant to ask about one type of contract and asked about another so their preliminary key was inconsistent with the text. THAT was later appealed and became invalid. (Why both weren't accepted, I don't know)

Extrapolating the past percentages of multiple answer or invalid multiple choice questions to essay questions, it may sometimes be possible that a candidate's score should have been two marks higher than what that candidate actually received.
Multiple choice attempt to be objective. Credit or No credit.
Grading the essays reflect partial credit, and it would be easy to get full credit if you simply specified your assumptions ("If by losses you mean the Model's losses"...) But that only works if you notice the problem (two interpretations) during the exam.

We should all be working towards a more open appeals process in which we receive all our answers so that we can defend our investments of time and emotional energy.Aaaaaah! :rofl: You just convinced me to switch my position and back the CAS on this one! Way to go!
The CAS exam committee are volunteers and their ability to handle appeals is limited. You've convinced yourself that two questions everyone else is clear on are obviously defective - I can imagine how many appeals you'd send for each partial credit you received.
Why not spend this boundless "defense" energy to study a little more and get just one or two more unambiguous questions right?

If I were designing the appeal process (i don't), I'd probably look at the # of passing candidates who answered a question correctly vs the alternative.
If PASSING candidates did not outscore the non-passing candidates because of multiple interpretations, then the question does not effectively test knowledge learned.

ramanujan
07-20-2004, 07:52 PM
Dear Math Guy and Ramanujan:

Okay, you two, help me out here. Is my reasoning correct?


Consider risks, A and B, for the Super Efficient Company with zero expenses and ALAE* (Zero expenses and ALAE are assumed to simplify the analysis, however the results apply more broadly, ... all profit is made on investment income).

Risk A has total premium and total Expected losses of 200,000, and basic limits premium (P) and basic limits expected losses (E) of 50,000.

Risk B has total premium and total expected losses of 100,000, and basic limits premium (P) and basic limits expected losses (E) of 100,000.

Consider the formula for the MSL

MSL = 0.3 * E / Z, where Z = P / (P + 100,000)

=> MSL = 0.3 * E (P + 100,000) / P

For Risk A, MSL = 0.3 * 50,000 (50,000 + 100,000) / 50,000 = 45,000
For Risk B, MSL = 0.3 * 100,000 ( 100,000 + 100,000) / 100,000 = 60,000

Clearly, Risk A is larger than Risk B, but since the basic limits portion of the risk is smaller, Risk A has a smaller MSL.

Therefore, Tiller should have said:
"The MSL increases as the size of the risk (on a basic limits basis) increases."


Are you in agreement with this, or have I missed something in my interpretation of Tiller?

What is your definition of "the size of risk"?

My belief is that Basic limits loss cost = size of the risk.

takeshire
07-20-2004, 08:25 PM
What is your definition of "the size of risk"?

My belief is that Basic limits loss cost = size of the risk.


According to Tiller, size of risk can be measured by premium, or exposure, or expected losses, or expected number of claims, or credibility.

Theoretically, very vague.

But by wording of the question, I would certainly not make the leap to think size of the basic limits portion of the risk.

Would you have honestly, before this conversation have said that my risk A is smaller than my risk B?

takeshire
07-20-2004, 08:37 PM
PROVE you are qualified (or more qualified than the rest), not that the CAS is a bad test writer. We already know that!


It's difficult to prove you are qualified if the measure you are using is not valid. Ambiguous questions are not valid measures of a candidate's being qualified, especially when the ambiguity cannot be spelled out and accounted for. I think the way we now measure qualification is like calculating the weight of jello using a protractor.

ramanujan
07-21-2004, 12:22 AM
What is your definition of "the size of risk"?

My belief is that Basic limits loss cost = size of the risk.


According to Tiller, size of risk can be measured by premium, or exposure, or expected losses, or expected number of claims, or credibility.

Theoretically, very vague.

But by wording of the question, I would certainly not make the leap to think size of the basic limits portion of the risk.

Would you have honestly, before this conversation have said that my risk A is smaller than my risk B?

If the basic Limits Loss cost is higher for risk B, why is the total limits Loss cost higher for Risk A?
Perhaps, because A is buying higher limits. Does that mean size of risk A is more, I don't think so.
Why would the basic limits loss cost be higher for a certain risk, if it is smaller size?

Utanapishtim
07-21-2004, 01:12 AM
If the basic Limits Loss cost is higher for risk B, why is the total limits Loss cost higher for Risk A?
Perhaps, because A is buying higher limits. Does that mean size of risk A is more, I don't think so.
Why would the basic limits loss cost be higher for a certain risk, if it is smaller size?
I'm stepping in late here and didn't pay much attention to the set-up, but it sounds like a pretty clear matter of high frequency versus high severity risk. For the simplest case, say Basic limits are $50k, then risk A is expected to have (and has) one loss of at least $200k, while risk B has two losses of only $50k each, while both buy total limits of $200k.

Or, I could be way off-base, if you're not talking about what I think you're talking about. As for me, exam 5 is behind me now, so I can forget about it now anyway (;

takeshire
07-21-2004, 01:30 AM
Ramanujan,

Utanaphishtim stated a clear example. Please think about the appeal and tell me if you can find any inconsistencies in it. I think it will be a strong appeal to point out that instead of "risk," Tiller should have said "the basic limits portion of the risk." This is consistent with my original intuition when answering the quesiton on the exam, that it is not just expected value of the risk that matters, but also variance. Actually, in this way it is clear that even an individual risk can have a decreasing MSL from one year to the next as the size of the risk increases, if the variance increases enough so that the basic limits portion of the risk decreases.


Thanks,
Takeshire

takeshire
07-21-2004, 03:35 PM
Aaaaaah! :rofl: You just convinced me to switch my position and back the CAS on this one! Way to go!
The CAS exam committee are volunteers and their ability to handle appeals is limited. You've convinced yourself that two questions everyone else is clear on are obviously defective - I can imagine how many appeals you'd send for each partial credit you received.
Why not spend this boundless "defense" energy to study a little more and get just one or two more unambiguous questions right?



I would disagree. I think that focusing my energy now on defense is the right place to spend my energy. Unless I have no case, of course.

Also, I don't think my viewpoint is as extreme as some of the reactions I'm receiving makes it appear. I merely proposed WORKING TOWARDS a more open process. I know it can't happen overnight.

Many of the nightmare scenarios, such as an incredibly swamped appeals process can be avoided by planning the pace of progress from an exam committee that spends too much time on the back end (appeals) because the front end (item writing) needs improvement, to an exam committee that works hard on the front end thus allowing an open appeals process that is not taxing on their time.

Writing (and peer reviewing) quality, unambiguous exam questions is not easy. It requires more than a deep understanding of the material, and a solid grasp of the language. It also requires a focus on the goal of unambiguity that approaches the artistic. I'm not sure how much, and what quality of training is currently provided to help item writers pursue that goal.

I lay a bet, however, that if the exam committee were to reduce the ambiguity of exam questions, they would WANT to open up the appeals process more, because they would actually get fewer appeals; candidates would clearly see the items that they could and could not appeal, and the bulk of "blind & desperate" appeals would be curtailed.


If I were designing the appeal process (i don't), I'd probably look at the # of passing candidates who answered a question correctly vs the alternative.
If PASSING candidates did not outscore the non-passing candidates because of multiple interpretations, then the question does not effectively test knowledge learned.


This is a reasonable suggestion, and I bet they already do it. However, I believe that the exam rewards rote learning and sometimes penalizes a broader interpretation of the material. The type of analysis you propose, while important, does not bridge the gap between the type of thinking we would like to see demonstrated, and the type of thinking that is actually rewarded on the exam.

In other words, 900 examinees have red lights, and 100 examinees have green lights. We want to know how often the the light is on (upstairs), but many of our items recognize that the light is on only if it is a red light. The analysis you propose would not tell the analyst that there may also be a minority of green lights out there that are on, since the exam itself is biased towards red lights. That is (an oversimplification of) the problem that I believe we face on the exam.

ramanujan
07-23-2004, 09:17 PM
If the basic Limits Loss cost is higher for risk B, why is the total limits Loss cost higher for Risk A?
Perhaps, because A is buying higher limits. Does that mean size of risk A is more, I don't think so.
Why would the basic limits loss cost be higher for a certain risk, if it is smaller size?
I'm stepping in late here and didn't pay much attention to the set-up, but it sounds like a pretty clear matter of high frequency versus high severity risk. For the simplest case, say Basic limits are $50k, then risk A is expected to have (and has) one loss of at least $200k, while risk B has two losses of only $50k each, while both buy total limits of $200k.

Or, I could be way off-base, if you're not talking about what I think you're talking about. As for me, exam 5 is behind me now, so I can forget about it now anyway (;

You are right in what you are saying but the CGL experience rating plan, that Tiller is talking about, does not care about the excess losses. All the losses that go in the plan are limited to basic limits. The plan does not look at frequency or severity, it just looks at the total expected losses below basic limits. So I do not know how can one acount be expected to have $200K of losses and other account two $50K of losses. For the plan it just means basic limits loss cost of 50K and 100K respectively.

Utanapishtim
07-24-2004, 01:07 PM
You are right in what you are saying but the CGL experience rating plan, that Tiller is talking about, does not care about the excess losses. All the losses that go in the plan are limited to basic limits. The plan does not look at frequency or severity, it just looks at the total expected losses below basic limits. So I do not know how can one acount be expected to have $200K of losses and other account two $50K of losses. For the plan it just means basic limits loss cost of 50K and 100K respectively.
Good ol' Tiller...

Still, in a generic sense you'd surely agree that referring to the size of a risk should account for its expected loss (maybe net of reinsurance) regardless of basic limits, right?

ramanujan
07-24-2004, 02:13 PM
Good ol' Tiller...

Still, in a generic sense you'd surely agree that referring to the size of a risk should account for its expected loss (maybe net of reinsurance) regardless of basic limits, right?

I agree, as long as the risk is large enough to have to have some credibility. But you have to still remember than the historical large losses would have lesser credibility than basic limit (or other attritional limit that you may choose) and above some limit (you mention reinsurance) the credibility would be zero.

takeshire
07-25-2004, 11:55 AM
I agree, as long as the risk is large enough to have to have some credibility. But you have to still remember than the historical large losses would have lesser credibility than basic limit (or other attritional limit that you may choose) and above some limit (you mention reinsurance) the credibility would be zero.


I'm not certain how one would calculate expected losses where there is no credibility. We're splitting hairs here, I think.

The fact that historical losses above basic limits have less credibility, is not a material detail in this question.

Ramanujan,

Can I get your input on something else?

Consider the equation:

MSL = 0.3E/Z
= 0.3 (BL losses)(D-ratio)/Z
= 0.3 (BL prem)(BL ELR)(D-ratio)(BL prem +100,000)/(BL prem)
= 0.3 (BL ELR)(D-ratio)(BL prem + 100,000)
= K(D-ratio)(BL prem) + K(D-ratio)100,000

Where BL = basic limits, and K = 0.3(BL ELR).

My understanding (which could be mistaken) is that the D-ratio is dependent on the MSL.

Considering Exhibit 1 in Tiller's study note, my understanding was that the MSL is plugged into the column on the far left to determine the D-ratio on the far right. But that would require some iterative process just to come up with the final D-ratio, and that certainly doesn't make sense as an algorithm to have every insurance company to go through as part of its filing. What is your take on this? How do you determine which D-ratio to use? I truly can't make heads or tails of how to apply the D-ratio by looking at the text and studying Exhibit 1.

Also, do you think the D-ratio would ever increase as the BL premium decreased? (It's difficult for me to see the implication of dependency on the direction of MSL as the BL premium increases. I imagine that the opposite could occur in some cases where the BL premium is small, but I'm not certain.)

I know this seems like splitting hairs, but I want to get it right on the appeal.


Thanks,
Takeshire

Colymbosathon ecplecticos
07-25-2004, 03:37 PM
Takeshire,

I'm glad that you are taking the time to learn this material, and I am confident that it will help you next time you take this exam.

I do have a question for you. When you send in your appeal, what exactly are you going to claim? That the question was unclear, but that your mastery of this topic was a such a high level that you could not determine what the question writer had in mind? I don't think that this thread bears that out. Do you want the question thrown out for some abstract reason that says that no question that isn't ideal should be allowed on the exams? Nobody likes a less than perfect question, but in the real world not every question is well-posed.

Again, good luck with your future exam progress.

takeshire
07-25-2004, 06:36 PM
Takeshire,

I'm glad that you are taking the time to learn this material, and I am confident that it will help you next time you take this exam.

I do have a question for you. When you send in your appeal, what exactly are you going to claim? That the question was unclear, but that your mastery of this topic was a such a high level that you could not determine what the question writer had in mind? I don't think that this thread bears that out. Do you want the question thrown out for some abstract reason that says that no question that isn't ideal should be allowed on the exams? Nobody likes a less than perfect question, but in the real world not every question is well-posed.

Again, good luck with your future exam progress.



You're absolutely right, Coly. I am still learning. And I have not completely mastered every portion of the material. Years from now, I'm sure I will still be looking back from time to time at old material and having the bulb light up again.

Regarding your question, I was surprised and glad to find after hashing the question out on this website that my intuition on item 17 was correct. Is intuition a fair mental tool to use on these multiple choice questions? If it is, then I think I deserve credit.

Now I have a question for you. Why didn't the statements I am appealing bother you, too? Question 16 is clearly invalid, and the text from which question 17 was quoted is clearly wrong. Perhaps it is because you didn't have your brain turned on when you were reading those questions during the exam. Or maybe you didn't have the material mastered.

Which was it? Brain off or lack of mastery of the material?

Or perhaps you read and understood what the question said, but then said to yourself, "I don't think the item-writer meant that, so I'll read the question the wrong way, thereby getting the right answer." If that was your line of thinking, I bow down to you in worship. :notworth: You are truly a superior intellect with your amazing analytical skills, your photographic memory, and your extra-sensory perception.


Takeshire

ramanujan
07-27-2004, 07:42 PM
I'm not certain how one would calculate expected losses where there is no credibility. We're splitting hairs here, I think.

Using the ILF/ELF/ELPPF.


Ramanujan,

Can I get your input on something else?

Consider the equation:

MSL = 0.3E/Z
= 0.3 (BL losses)(D-ratio)/Z
= 0.3 (BL prem)(BL ELR)(D-ratio)(BL prem +100,000)/(BL prem)
= 0.3 (BL ELR)(D-ratio)(BL prem + 100,000)
= K(D-ratio)(BL prem) + K(D-ratio)100,000

Where BL = basic limits, and K = 0.3(BL ELR).

My understanding (which could be mistaken) is that the D-ratio is dependent on the MSL.

Considering Exhibit 1 in Tiller's study note, my understanding was that the MSL is plugged into the column on the far left to determine the D-ratio on the far right. But that would require some iterative process just to come up with the final D-ratio, and that certainly doesn't make sense as an algorithm to have every insurance company to go through as part of its filing. What is your take on this? How do you determine which D-ratio to use? I truly can't make heads or tails of how to apply the D-ratio by looking at the text and studying Exhibit 1.

Also, do you think the D-ratio would ever increase as the BL premium decreased? (It's difficult for me to see the implication of dependency on the direction of MSL as the BL premium increases. I imagine that the opposite could occur in some cases where the BL premium is small, but I'm not certain.)

I know this seems like splitting hairs, but I want to get it right on the appeal.


Thanks,
Takeshire

If you adopt the ISO CGL experience rating plan without modification, you do not need to calculate the D-ratio and MSL. It is already calculated by ISO.

D-ratio increases as MSL increases and MSL increases as Basic Limits Premium increases => D-ratio increases as BL Premium increases. You can also see this relationship in the tables which are included on the ISO CGL experience rating plan.

takeshire
07-29-2004, 11:38 AM
Thanks, Ramanujan. Any time you need assistance with an appeal, just email me at takeshire@yahoo.com