Sick of ULAE

[Officer Sun]

Does anyone other than me find it a little unreasonable that the ULAE chapter has about as many methods as all of the other chapters combined? This is an especially aggravating last chapter to try and get through. I also find this a little ironic when ULAE is something that is often largely ignored in practice.

[jesusislord]

it is almost unbearable to read and decipher what is going on. however, tia videos make it so simple and understandable. if you have tia, i would just watch the vids and skip the reading.

[Officer Sun]

The cost of TIA has become prohibitive, especially if you are only taking half the exam. I don't think it is worth the money given the level of difficulty of this exam.

[TheZaha]

Thanks for the tip JIL

[Vorian Atreides]

For starters on this chapter, I would focus on understanding the difference between the dollar-based techniques and the count-based techniques (e.g., the assumptions each are based on and how they're different in how ULAE dollars are allocated wrt reserve estimates).

Then, based on past Exam problems:

Then work to understand and master the traditional, Kittel's refinement, and the generalized Kittel's (Conger & Nolibos) methods of the dollar-based techniques. I know that the notation here seems unweildy, but focus on the qualitative aspects of the generalized method wrt what's going on and I think you'll begin to see how to remember the other two methods pretty well.

As for the count-based methods, Look at the "Brian" technique and Wendy Johnson's methods--assumptions and weaknesses (or shortcomings) of each.

I wouldn't worry about the rest until you get these down pretty good, IMO.

[Officer Sun]

Thanks for the tip, but I think it is still overly time intensive no matter how you slice it up. I think that Freidland has done a really good job streamlining this material, but I wish that she would have cleaned this chapter up a little more like all of the rest. In the unlikely event that anyone ever cares about ULAE they can always do the research to find all the variations that are out there.

[Elsaball]

Does anyone other than me find it a little unreasonable that the ULAE chapter has about as many methods as all of the other chapters combined? This is an especially aggravating last chapter to try and get through. I also find this a little ironic when ULAE is something that is often largely ignored in practice.

It isn't used in practice? Really? We spend a bit of time on ULAE.

[Scars]

It isn't used in practice? Really? We spend a bit of time on ULAE.

Ditto.

[General Apathy]

It isn't used in practice? Really? We spend a bit of time on ULAE.

Ditto.

It's often used in practice. the question is should it be?

Is there a real use for ULAE?

[Vorian Atreides]

It's often used in practice. the question is should it be?

Is there a real use for reserving ULAE?
IATYP . . .

On the pricing side, it needs to be accounted for.

On the reserving side, the question is valid to the extent of addressing: how much ULAE expenses will be needed (that is currently unpaid) to settle all outstanding claims if the company were to cease operations today?

[Vorian Atreides]

Quick Update:

As for the dollar-based estimation, it's not too difficult in keeping the different approaches aligned.

The main difference between the methods seems to be in how the ULAE ratio is calculated.

More specifically, W* is the selected ratio to apply, and it's based on yearly calculations of W_i = M_i / B_i where

W_i = ULAE ratio for year i (see note below regarding i),
M_i = ULAE actually paid out during year i, and
B_i = Basis (Losses)

Note that i is an index over a given (fixed) time interval--like a CY or quarter, etc.

And it seems that M_i is identical in all methods, but that it's B_i that's different for each of the methods. Using the Generalized method as the starting point, we see that B_i is a weighted average:

B_i = U1*R_i + U2*P_i + U3*C_i

Where
U1 + U2 + U3 = 1.000 (these are the weights)
R_i = Reported Claims (ultimate cost) during i
P_i = Paid Claims during i
C_i = Closed Claims (ultimate cost) during i

And the generalized method basically calculates the total ULAE need and then backs out the incurred ULAE during i.

The Classical (or Traditional) and Kittel Refinement methods don't estimate the ultimate costs of claims for the basis (key distinction here), but uses the same general formula for the basis with these specific values for the weights:

Classical (Traditional): U1 = U3 = 0%; U2 = 100%
Kittel Refinement: U1 = U2 = 50%; U3 = 0%

(Note that there is a simplified Generalized method--U3 = 0%, but U1 and U2 are not fixed.)

So, if given actual paid & reported losses as well ultimate paid & reported losses, you would use the latter two only for the Generalized method; you would use the former for the other two methods.

Now, as for the Mango-Allen refinement, the basis is calculated based on expected claims rather than actual (or developed) claims. Note that this method is useful for a new company or highly volatile lines. The text does point out that this method isn't as efficient for more established lines/company.

[timxu]

Good post VA. I reached a similar conclusion. The only note is that the definition of reported loss is a bit different between kittel and the general method - kittel uses cy reported loss while the conger method uses "ultimate loss of claims reported during a year".

I think the book would be more clear if it presented methods in opposite order and focused on how each next method a derivative of the prior.

[Vorian Atreides]

I agree that the material could be presented more clearly. However, I think it's a matter that Friedland isn't an expert in this particular field and she borrowed heavily from the Conger & Nolibos paper (there is a lot of verbatim statements from this paper in Ch 17).

[Revis Island]

Well this is precisely the problem with the chapter. It is a total laundry list of methods in a very raw and unfinished form. I seem to recall that the very first version of Friedland did not originally include Chapter 17. It said that the ULAE section would be added later, further confirming that the chapter was an afterthought. This is my second time through exam 5 and ULAE in Chapter 17 of Friedland is my most dreaded topic. I can never keep all the methods straight. But, thanks VA for the summary. I will use it as I try and attack chapter 17 today.

[Vorian Atreides]

I believe that Chapter 17 wasn't included in the first version of Friedland because they were trying to get (or negotiate) the permissions to utilize significant portions of the Conger & Nolibos verbatim w/o the standard citations.

IIRC, the first time Ch 17 did appear in Friedland, it was before a sitting for old Exam 6, so the Conger paper remained on the syllabus for that sitting (and was removed soon afterward).

[asile]

Quick Update:

As for the dollar-based estimation, it's not too difficult in keeping the different approaches aligned.

The main difference between the methods seems to be in how the ULAE ratio is calculated.

More specifically, W* is the selected ratio to apply, and it's based on yearly calculations of W_i = M_i / B_i where

W_i = ULAE ratio for year i (see note below regarding i),
M_i = ULAE actually paid out during year i, and
B_i = Basis (Losses)

Note that i is an index over a given (fixed) time interval--like a CY or quarter, etc.

And it seems that M_i is identical in all methods, but that it's B_i that's different for each of the methods. Using the Generalized method as the starting point, we see that B_i is a weighted average:

B_i = U1*R_i + U2*P_i + U3*C_i

Where
U1 + U2 + U3 = 1.000 (these are the weights)
R_i = Reported Claims (ultimate cost) during i
P_i = Paid Claims during i
C_i = Closed Claims (ultimate cost) during i

And the generalized method basically calculates the total ULAE need and then backs out the incurred ULAE during i.

The Classical (or Traditional) and Kittel Refinement methods don't estimate the ultimate costs of claims for the basis (key distinction here), but uses the same general formula for the basis with these specific values for the weights:

Classical (Traditional): U1 = U3 = 0%; U2 = 100%
Kittel Refinement: U1 = U2 = 50%; U3 = 0%

(Note that there is a simplified Generalized method--U3 = 0%, but U1 and U2 are not fixed.)

So, if given actual paid & reported losses as well ultimate paid & reported losses, you would use the latter two only for the Generalized method; you would use the former for the other two methods.

Now, as for the Mango-Allen refinement, the basis is calculated based on expected claims rather than actual (or developed) claims. Note that this method is useful for a new company or highly volatile lines. The text does point out that this method isn't as efficient for more established lines/company.

More color on VA's excellent post:

Another way to think of it (perhaps more consistent with the definitions of U1 = opening (ultimate on claims reporting), U2 = maintaining (claims paid), U3 = closing (ultimate on claims closed)) is that under the assumption of no partial losses, the classical method is U3 = 100% (because any payment is a close) and U1=U2 = 0%.

If you are willing to assume that prior years' case reserves are totally accurate, you can make the same jump for Kittel: U1 = 50% because CY Incurred = Ultimate on Reported Claims (payouts on prior years' losses will be netted against reserve takedowns); U3= 50% per above.

[Vorian Atreides]

More color on VA's excellent post:

Another way to think of it (perhaps more consistent with the definitions of U1 = opening (ultimate on claims reporting), U2 = maintaining (claims paid), U3 = closing (ultimate on claims closed)) is that under the assumption of no partial losses, the classical method is U3 = 100% (because any payment is a close) and U1=U2 = 0%.

If you are willing to assume that prior years' case reserves are totally accurate, you can make the same jump for Kittel: U1 = 50% because CY Incurred = Ultimate on Reported Claims (payouts on prior years' losses will be netted against reserve takedowns); U3= 50% per above.
The bolded is equivalent to what I had posted for the classical (aka traditional) method because in this case: C_i = P_i.

[Revis Island]

Haha, General Apathy, I just watched the Big Lebowski for the first time two nights ago! I also just spent all day trying to wrap my head around all the different ULAE methods. I'm sick of this. Maybe I'm just a little slow, but I find this to be the most difficult concept out of Werner/Friedland combined. It just seems like there are multiple ways for them to obscure the pieces that you need and if you don't have a solid understanding of what pieces you are using you'll be screwed. I finally gave up and went straight to the Conger/Nolibos paper which I actually found easier to read than the Friedland summary of the Conger/Nolibos paper. I have also had a print out of VA's summary which was very helpful. Thanks for that.

[Vorian Atreides]

Glad it helped. :toth:

[Vypa]

The claims basis, B, in the Generalized approach is computed on a Calendar Year basis, correct to calculate W*, where W* = CY Paid ULAE / B.

However, it appears from the text that the B used to actually estimate unpaid ULAE via either the 'Bornhuetter-Ferguson' or 'Development' Method should be an Accident Year metric since it's compared to the estimated ultimate losses for the accident years, L.

Is this correct?