PDA

View Full Version : Jones: Two-step trending


Examinator
02-04-2005, 02:26 PM
In Jones' paper on trend, he discusses both a one-step and a two-step trending procedure. The one-step trends from the average written date of the experience period to the average written date of the future policy period. No problems here. The two-step does the same thing, except splitting the trending period at the average date for the latest trend point. Can someone explain what this point is?

He describes the first step factor as the average on-level written premium from the latest year divided by the average on-level earned premium for the specific year. What does this represent? Where does the average date for the latest trend point come in? It seems to me that the two-step method simply adds an invented step that's not necessary, but I'm obviously confused somewhere. Thanks for your help.

J.T.
02-04-2005, 02:57 PM
Let me see if I can help. I haven't gotten to his article yet, but I'll try to redigest what I remember.

The point to break the first step of trending is better understood (at least by me) like this: Say you are trending 3 years worth of data (01, 02, 03). You want to trend this data to the future effective date of 1/1/05. Take the last period you are trending (03) and pick the midpoint of it (7/1). That is where you end the first step of trending at.

As for the factor, that is to get the EP at the same level of WP as current. IOW, if in 01 we didn't write as much WP as we have in 03 (separate from CRL), then this will get it on-level in that aspect.

I think of it this way. You may need trend for two different time periods: The time during when the historical data was collected, and the time for the period from the historical data to the time the new rate change goes into effect.

Hope this helps, but I may have muddied the waters more for you. I should cover this in a week or 2, so I will be able to help more then.

jk
02-04-2005, 03:00 PM
The two-step does the same thing, except splitting the trending period at the average date for the latest trend point. Can someone explain what this point is?Splitting the trending period recognizes that a trend which you expect to operate smoothly in the future may not have done so in the recent past.

He describes the first step factor as the average on-level written premium from the latest year divided by the average on-level earned premium for the specific year. What does this represent?It effectively reprices the earlier exposures at the most current price available--that is, the most recent average written premium per exposure. Where does the average date for the latest trend point come in?That's where you split the trend. The most recent average written premium per exposure is "as of" the average written date of the most recent experience period. You use Step 1 to reprice the earlier exposures at that level, and then use Step 2 to trend from that date to the average date of the forecast period.

Examinator
02-05-2005, 08:02 AM
Let me see if I have this down. The break point is the midopoint of the latest experience period, also the average written date of the latest experience period.

The factor simply adjusts the on-level premium from each experience period to the latest on-level written premium of the latest experience period.

The second factor, which I think I understand, trends out to the average written date of the exposure period.

jk
02-09-2005, 11:31 PM
Sounds right to me.

great3981
02-11-2005, 08:47 AM
The break point is not necessarily the midpoint of the latest experience period. Jones uses a 12 month moving average of the premiums, so the midpoint of that latest average is 7/1. If you were trending 6 month policies, however, your break point does not change as Jones still uses a 12 month moving average.

Consult page 17 in Jones for what may be a better explanation.

Examinator
02-19-2005, 09:07 AM
That's where you split the trend. The most recent average written premium per exposure is "as of" the average written date of the most recent experience period.

I understand that the factor in the first step (which is obtained for each CY by the latest average written premium divided by the CY average earned premium) brings the total earned premium for each CY to the level of the latest total written premium. But this date, the midpoint of the latest experience period in Jones, is not the average written date of the most recent experience period (is it?). The average written date of the latest experience period, which is CY 2001, would be 1/1/2001, not 7/1/2001. This date stays the same regardless of policy length.

I'm just trying to associate in my head what that date truly is. For a while I thought it was an average expiration date of the CY in question, but that's not right. It seems like it should be the average written date of the latest period, give how the factor is derived, but wouldn't that be 1/1/2001?

jk
02-19-2005, 11:33 PM
But this date, the midpoint of the latest experience period in Jones, is not the average written date of the most recent experience period (is it?). The average written date of the latest experience period, which is CY 2001, would be 1/1/2001, not 7/1/2001.Why do you say that? Under the standard assumption that policies are written uniformly throughout the year, the average written date for policies written during 2001 is 7/1/2001.

Examinator
02-22-2005, 09:25 AM
My logic is that the average earned date for policies uniformly written in CY 2001 is 7/1/2001. The average written date is half the policy period prior, or 1/1/2001. Isn't this what Jones says for his CY 2000 trending example?

jk
02-22-2005, 11:34 AM
The average written date for policies earned during 2000 is 1/1/2000. The average written date for policies written during 2001 is 7/1/2001.

Jones (in Exhibit 5) is trending a series of historical earned premiums based on trend factors deduced from a series of average written premiums. To trend premium earned during 2000, he first brings it to the average level of premium written during 2001--that is, 18 months of trend. Then, in Step 2, he trends it out to policies that will be written during 2003--two more years of trend.

In Exhibit 6, he considers the possibility of trending a series of historical earned premiums based on trend factors deduced from a series of average earned premiums. Then Step 1 is 12 months and Step 2 is 30.

Examinator
02-22-2005, 12:04 PM
Finally I see the error in my thought process. I wasn't distinguishing averages dates for written/earned policies in a given CY. This makes sense to me now. Thanks for your help and patience.

Nut
03-27-2006, 10:44 AM
Hi all,

I'm still quite confused by this issue. Here is my current understanding.

Assume we have WP information as in Exhibit 5.

For Step 1:

Beginning point:
Use CY Earned premium, and use the average Writing Date for this premium. This sort of "converts" the EP on a WP basis? If we already have WP info, why do we even bother with the EP if Jones has established that using WP is preferred?

Ending Point:
Use Policy Year Average Written premium, and use the corresponding Average Writing date for this premium. Is this what Jones calles the "midpoint of the latest data point"?

For Step 2:

Beginning Point:
Use ending point of Step 1.

Ending Point:
Use Policy Year Average Written Premium, and use the corresponding Average Writing Date for this Premium.

---------------

What is wrong with my logic? Looks like Jones is mixing the use PY and CY quite a bit! He diagrams the end point of the second step on a PY basis, but All of Step 1 appears to be on a CY basis.

Nut
03-27-2006, 10:48 AM
Appendix 2, pg 22, column (3) of Jones shows for step 1:

(Latest Avg WP @ CRL) / (Avg CY EP @ CRL)

Mahler's solution uses for step 1:
(Latest Avg WP @ CRL) / (Avg CY WP @ CRL) .

I did the problem by using the WP rates and earning them into the CYs. Of course, I got problem wrong.

I'd appreciate if anyone could clarify exactly what we are supposed to use for each of the points in Step 1 and Step 2 of a two step procedure. My previous post outlines my confused understanding.

Thanks!
NUT

frank_exams
03-27-2006, 06:10 PM
Let me try to offer some help.

For Step 1:

Beginning point:
Use CY Earned premium, and use the average Writing Date for this premium. This sort of "converts" the EP on a WP basis? If we already have WP info, why do we even bother with the EP if Jones has established that using WP is preferred?


Whoops! I misread this part of your post earlier. I'm not sure what the reasoning is, other than for examiners to ask us for specific CAY trend factors. in the first step. I agree it's kind of redundant if you already have WP.


Ending Point:
Use Policy Year Average Written premium, and use the corresponding Average Writing date for this premium. Is this what Jones calles the "midpoint of the latest data point"?


Note that you're trending a 12-mo moving avg WP, i.e. you use the previous 12 months worth of WP just to get one data point. The first step actually ends at the midpoint of the last data point. The last data point would be the last 12 months of the experience period, so natch, the first step ends 6 months before the end of the experience period.


For Step 2:

Beginning Point:
Use ending point of Step 1.

Ending Point:
Use Policy Year Average Written Premium, and use the corresponding Average Writing Date for this Premium.

---------------

What is wrong with my logic? Looks like Jones is mixing the use PY and CY quite a bit! He diagrams the end point of the second step on a PY basis, but All of Step 1 appears to be on a CY basis.

1-step and 2-step are different. In 1-step, you're using CAY premium, which is typically earned. In the 2-step, you're looking at 12-mo moving average WP. The WP used in the analysis begins 12 months prior to the CAY you're analyzing and ends the last day of the CAY. It just so happens that the average written date turns out to be the same as in the 1-step version.

The end point for both is the average written date for the future period on a PY basis because you're no doubt performing this analysis for a proposed rate change. To use CAY would sully things by implicitly including premium earned at current rate level into the future exposure period.

Frank

frank_exams
03-27-2006, 06:24 PM
Appendix 2, pg 22, column (3) of Jones shows for step 1:

(Latest Avg WP @ CRL) / (Avg CY EP @ CRL)

Mahler's solution uses for step 1:
(Latest Avg WP @ CRL) / (Avg CY WP @ CRL) .

I did the problem by using the WP rates and earning them into the CYs. Of course, I got problem wrong.

I'd appreciate if anyone could clarify exactly what we are supposed to use for each of the points in Step 1 and Step 2 of a two step procedure. My previous post outlines my confused understanding.

Thanks!
NUT

I think both may be right, depending on what data (EP or WP) you're given to start with. For example, 2005Q37 gives you WP. As you probably know the first step of the 2-step process is a pretty lame, you're just converting CAY premium to the latest CAY avg WP. So the factor will always be:

(Latest avg WP @ CRL) / (whatever they give you)

Frank

Nut
03-28-2006, 12:46 PM
Thanks Frank.

For Jones 2005Q37, wouldn't it be more conistent with Jones to use the earn the WP given into the CAY in question? For example, half of the average WP of $1000 in CY 2002 would get Earned in 2002, and half would be earned in 2003. Likewise, half of the Ave WP of 933.33 in CY 2003 would get Earned in 2003, and half would get earned in CY 2004. From here, I used the Earned premiums in each CY to figure out the EP @ Current Rate Level (ie 2004 Rate Level).

Of course I got the question wrong as usual. Perhaps I should just gloss over this and do what you suggest below, (Latest avg WP @ CRL) / (whatever they give you), and not worry about earning premiums!

Thanks again.

frank_exams
03-28-2006, 05:15 PM
For Jones 2005Q37, wouldn't it be more conistent with Jones to use the earn the WP given into the CAY in question? For example, half of the average WP of $1000 in CY 2002 would get Earned in 2002, and half would be earned in 2003. Likewise, half of the Ave WP of 933.33 in CY 2003 would get Earned in 2003, and half would get earned in CY 2004. From here, I used the Earned premiums in each CY to figure out the EP @ Current Rate Level (ie 2004 Rate Level).

It's frustrating when an apparently correct solution may not be right, especially since the approach seems good and the data given is inconsistent.

One problem with adjusting to EP is you're missing 2001 data, so you won't have the complete EP figure for 2002. I'd just stick with the slap the data into the denominator strategy. The irony of this is that, knowing me, I'll probably make a careless mistake and miss this problem on the exam.

Thanks again.

Glad I could be of help.

Frank

WhitewaterGirl
03-29-2006, 08:54 AM
Frank, There is no doubt in my mind that you're going to pass this exam. You're always all over these questions...

Utopial
04-23-2006, 09:53 PM
Note that you're trending a 12-mo moving avg WP, i.e. you use the previous 12 months worth of WP just to get one data point. The first step actually ends at the midpoint of the last data point. The last data point would be the last 12 months of the experience period, so natch, the first step ends 6 months before the end of the experience period.

for 6 month policies, isnt the last data point be the last 6 months of the experience period?
e.g. if u'r looking at EP for 2000, last data point are the policies written 12/31/00 - 6/30/01 and the midpoint of that is 1/4 of the way into 01...
this would mean that the middle point in 2 step trending is different for 6 and 12 month policies. can u point out where i'm going wrong?
thanks

2M
04-23-2006, 10:04 PM
for 6 month policies, isnt the last data point be the last 6 months of the experience period?
e.g. if u'r looking at EP for 2000, last data point are the policies written 12/31/00 - 6/30/01 and the midpoint of that is 1/4 of the way into 01...
this would mean that the middle point in 2 step trending is different for 6 and 12 month policies. can u point out where i'm going wrong?
thanks

According to the reading (page 17)

The "last data point" (break point) is the same for both 6 month and 12 month policies.

It is the 1st step and last step that are different.

The article cites the reason as 12-month moving averages of written premium are used in both cases.

Sox34
04-23-2006, 10:39 PM
It is the 1st step and last step that are different.


Actually, I believe it is that only the first step is different. The breakpoint and second step are the same for 6 and 12 month policies. This is because the avg written date in the future period is not dependent on the policy length.

The overall trend period for 6 month policy ends up being 3 months shorter.

Utopial
04-23-2006, 11:50 PM
According to the reading (page 17)

The "last data point" (break point) is the same for both 6 month and 12 month policies.

It is the 1st step and last step that are different.

The article cites the reason as 12-month moving averages of written premium are used in both cases.

wat's the point of using 12 months rolling avg? why not just trend it to the real midpoint of the last data point?

also, y does the 2005 jones question trend to the midpoint of the last period in the series (2004) rather than the midpoint of the last data point (last data point being the policy written 12/31/2004 with a midpoint of half way thru 2005?

maybe my definition of 'last data point' is wrong
im taking it to mean the term of the very last policy written in the experience period

2M
04-24-2006, 12:10 AM
Actually, I believe it is that only the first step is different. The breakpoint and second step are the same for 6 and 12 month policies. This is because the avg written date in the future period is not dependent on the policy length.

The overall trend period for 6 month policy ends up being 3 months shorter.

:oops: I should have kept reading when I was quoting the text

Utopial
04-24-2006, 12:19 AM
Actually, I believe it is that only the first step is different. The breakpoint and second step are the same for 6 and 12 month policies. This is because the avg written date in the future period is not dependent on the policy length.

The overall trend period for 6 month policy ends up being 3 months shorter.

for written the 2nd period is the same
for earned the 2nd period wouldnt be the same cause the avg earned end point is different

im still confused about this break point thing. what exactly is the 'last data point'??

2M
04-24-2006, 12:21 AM
wat's the point of using 12 months rolling avg? why not just trend it to the real midpoint of the last data point?

also, y does the 2005 jones question trend to the midpoint of the last period in the series (2004) rather than the midpoint of the last data point (last data point being the policy written 12/31/2004 with a midpoint of half way thru 2005?

maybe my definition of 'last data point' is wrong
im taking it to mean the term of the very last policy written in the experience period

Actually, the 2005#37 Jones problem does trend the breakpoint to the midpoint of the last period. The period is 2004. What it does is it trends to the average written date(since we are dealing with written premium)

The average written data in 2004 is the midpoint or 7/1/2004. All policies are annual, thus the average earned date of the average policy is 6 months later or 12/31/2004

2M
04-24-2006, 12:25 AM
for written the 2nd period is the same
for earned the 2nd period wouldnt be the same cause the avg earned end point is different

im still confused about this break point thing. what exactly is the 'last data point'??

Sox34 is correct

Quote from Jones

In discussing 6 month policies
"The second step results in the same length trending period as was used for 12-month policies, because the average written data in the future policy period does not depend on the length of the policies"

2M
04-24-2006, 12:28 AM
im still confused about this break point thing. what exactly is the 'last data point'??

What you are referring to as the 'last data point' and what we are calling the break point are probably the same thing.

As in 2005#37, your experience period is 2002, 2003 & 2004.

Your first step is to trend all the years in your experience period up to the llast point in your experience period (in this case 2004)

Your second step is to trend from that point up to the midpoint of your experience period (in this case 2006)

What you are calling the 'last data point' (in this case 2004), the text calls 'break point'

Utopial
04-24-2006, 12:32 AM
isnt the last data point the policy written on 12/31/2004 with a midpoint of half way thru 2005?

Utopial
04-24-2006, 12:34 AM
Sox34 is correct

Quote from Jones

In discussing 6 month policies
"The second step results in the same length trending period as was used for 12-month policies, because the average written data in the future policy period does not depend on the length of the policies"

yeh but jones is only talking about written periods

2M
04-24-2006, 12:35 AM
yeh but jones is only talking about written periods

yes, I was just going to post a correction, you are right

Utopial
04-24-2006, 12:37 AM
What you are referring to as the 'last data point' and what we are calling the break point are probably the same thing.

As in 2005#37, your experience period is 2002, 2003 & 2004.

Your first step is to trend all the years in your experience period up to the llast point in your experience period (in this case 2004)

Your second step is to trend from that point up to the midpoint of your experience period (in this case 2006)

What you are calling the 'last data point' (in this case 2004), the text calls 'break point'


exhibit 5 seems to contradict q37 2005. ex5 trends to the midpoint of the last policy, whereas q37 trends to the avg written date of the last period in the series

2M
04-24-2006, 12:37 AM
isnt the last data point the policy written on 12/31/2004 with a midpoint of half way thru 2005?

I think you might be confusing 'last data point' with last trend point in the period.

When you do trending, you don't trend from the last data point, you trend from the average point

2M
04-24-2006, 12:41 AM
exhibit 5 seems to contradict q37 2005. ex5 trends to the midpoint of the last policy, whereas q37 trends to the avg written date of the last period in the series

Really? I don't see that, it says "average date for latest trend point" is 7/1/01

Utopial
04-24-2006, 12:48 AM
Really? I don't see that, it says "average date for latest trend point" is 7/1/01

all10 says that hypothetically in the initial comments: 'IF the latest trend point in the series is for the yr ending 12/31/01, then.... 7/1/01'

if u go down the 'problem specific' ull see: 'the ratios are the trend factors for step 1. they are used to trend the premiums to 7/1/04' ie the end of step 1 is 7/1/04 which is the middle of the last yr in the series of the exper period

based on my retarded logic, ex5, ex7 and 2005#37 all contradict one another. the only thing that they all seem to do in common is trend to the middle of a 12 month period. ex5&7 trend to the middle of the year following the expereince period. q37 trends to the middle of the last yr of the experience period

2M
04-24-2006, 12:57 AM
all10 says that hypothetically in the initial comments: 'IF the latest trend point in the series is for the yr ending 12/31/01, then.... 7/1/01'

if u go down the 'problem specific' ull see: 'the ratios are the trend factors for step 1. they are used to trend the premiums to 7/1/04' ie the end of step 1 is 7/1/04 which is the middle of the last yr in the series of the exper period

based on my retarded logic, ex5, ex7 and 2005#37 all contradict one another. the only thing that they all seem to do in common is trend to the middle of a 12 month period. ex5&7 trend to the middle of the year following the expereince period. q37 trends to the middle of the last yr of the experience period


The experience periods in exhibit 5-7 are all calendar/accident years, as is q37. They all trend to the midpoint of the last year. NOT the year following.

Utopial
04-24-2006, 12:59 AM
but in ex5&7 no policies are written in 2001, so how is it an experience period?

2M
04-24-2006, 01:05 AM
I am not sure I follow:


The first line of the exhibit says 'Experience Period = 1999 through 2001'


They are specifically showing us an example for Calendar Accident Year 2000

2M
04-24-2006, 01:10 AM
A good example that might help you is 2004#35

Utopial
04-24-2006, 04:18 AM
thanks, i checked it out. it's pretty much just like the notes

i do have a question for you though that may solve my dilemma:
just say u have an 'experience period' of 2000-2010 and you are looking at CAY 2002. is the latest trend point the middle of 2010?
if so, how is the experience period chosen?

SouthOfSanity
04-24-2006, 11:46 AM
thanks, i checked it out. it's pretty much just like the notes

i do have a question for you though that may solve my dilemma:
just say u have an 'experience period' of 2000-2010 and you are looking at CAY 2002. is the latest trend point the middle of 2010?
if so, how is the experience period chosen?
I assume "latest trend point" means the breakpoint in a two-step trending. I believe the latest trend point is based on the average premium trend series. The time period may be unrelated to the experience period we are using to calculate historical loss ratios for the indication calculations. When we have an experience period of 2000-2010, I believe in Jones' terms it refers to the trend series, not necessarily the experience period we use to calculate loss ratios.

At least that is how I understand it. I hope this is not adding to your confusion.

Utopial
04-24-2006, 11:59 AM
so what would the trend series be in this case?
premiums written from 1/1/01 - 31/12/02
and thus earned 1/1/01 - 31/12/03?

Utopial
04-24-2006, 12:04 PM
2005q37 looks at avg written prem in 02,03 and 04. premium written in 04 would be earned over 05 as well.
if the break point is based on the trend series, then the break point would be 7/1/05 instead of the 7/1/04 that the solution uses.
if the break point is based on the experience period, then it would be 7/1/04

SouthOfSanity
04-24-2006, 12:38 PM
2005q37 looks at avg written prem in 02,03 and 04. premium written in 04 would be earned over 05 as well.
if the break point is based on the trend series, then the break point would be 7/1/05 instead of the 7/1/04 that the solution uses.
if the break point is based on the experience period, then it would be 7/1/04
I don't understand why you would want to earn the premiums out to 2005. The trend series ends on 12/31/2004, and we are trending the written premiums because Jones prefers written over earned in the determination of premium trends. So the latest average trend date for the trend series is 7/1/2004. This question also assumes CAY 2004 as the latest year in the experience period used to calculate historical loss ratios, though we do not need to know this information to solve the problem.

2M
04-24-2006, 12:43 PM
We are not trending to '05 in this question for step 1, the experience period runs only through Calendar year '04

SouthOfSanity
04-24-2006, 12:46 PM
so what would the trend series be in this case?
premiums written from 1/1/01 - 31/12/02
and thus earned 1/1/01 - 31/12/03?
The trend series would be what the examiner tells us. We don't have to earn anything or manipulate any numbers to decipher the trend series. If the analysis to determine the trend uses data from 1/1/01 to 12/31/02, then the trend series is 1/1/01-12/31/02. Jones prefers written to earned premium data to be used in the determination of the premium trend, so we do not earn the premiums.

TRINIDON2K
04-24-2006, 12:49 PM
i learned the 2 step from a 50 cent video...i practice in the clubs

Utopial
04-24-2006, 01:05 PM
I don't understand why you would want to earn the premiums out to 2005. The trend series ends on 12/31/2004, and we are trending the written premiums because Jones prefers written over earned in the determination of premium trends. So the latest average trend date for the trend series is 7/1/2004. This question also assumes CAY 2004 as the latest year in the experience period used to calculate historical loss ratios, though we do not need to know this information to solve the problem.

a policy written on 12/31/2004 is earned over 2005. comparing this to exhibit 5, the break point should be midway thru 05

Utopial
04-24-2006, 01:09 PM
We are not trending to '05 in this question for step 1, the experience period runs only through Calendar year '04

so do u look at experience period or trend series?
just say u have an 'experience period' of 2000-2010 and you are looking at CAY 2002. is the latest trend point the middle of 2010 or the middle of 2003?

2M
04-24-2006, 01:16 PM
so do u look at experience period or trend series?
just say u have an 'experience period' of 2000-2010 and you are looking at CAY 2002. is the latest trend point the middle of 2010 or the middle of 2003?

One way to think of it is the latest trend point in this example would be 7/1/10

SouthOfSanity
04-24-2006, 04:16 PM
a policy written on 12/31/2004 is earned over 2005. comparing this to exhibit 5, the break point should be midway thru 05
Again, we do not earn the premiums when we determine the latest average trend date (breakpoint) in step 1. The experience period used for loss ratios and the trend series are two different things. If you look at Exhibit 6, the breakpoint is the same regardless of whether you are using average written date for earned premium or average earned date as your trending date basis. This is because the trend series is a 12-month moving average series. You do not earn anything, it's just a snapshot at each data point.

When we determine the beginning trend date and the step 2 ending trend date, yes we do look at the earning of the premiums because that is what we are ultimately trending -- we are trending the premiums in our experience used to calculate the loss ratios. But we are using the annual trend that is determined from a trend series.

Step 1 in the two-step trending trends the experience premium to the latest available trend data date. Step 2 in the two-step trending projects the experience premium from the latest available trend data date to the projected future experience period.

Utopial
04-25-2006, 12:53 AM
ok guys, thanks a heap for the help. i really appreciate it
i think my misunderstanding was with regards to the experience period or latest available trend data.
good luck on the exam

Utopial
04-25-2006, 11:46 AM
btw - im pretty sure that the way the 2003 premium is adjusted in the solution for the rate decrease is incorrect

solution says: 0.5*933 + .5*933*.8 = 840

my solution:
x*1/2 + x *1/2*.8 = 933
solve for x
on level prem = x*.8 = 829

2M
04-25-2006, 12:15 PM
btw - im pretty sure that the way the 2003 premium is adjusted in the solution for the rate decrease is incorrect

solution says: 0.5*933 + .5*933*.8 = 840

my solution:
x*1/2 + x *1/2*.8 = 933
solve for x
on level prem = x*.8 = 829


You are correct

I get 829 as well. With that you have to 'select' the trend to arrive at the CAS solution.

pcact
04-25-2006, 05:57 PM
btw - im pretty sure that the way the 2003 premium is adjusted in the solution for the rate decrease is incorrect

solution says: 0.5*933 + .5*933*.8 = 840

my solution:
x*1/2 + x *1/2*.8 = 933
solve for x
on level prem = x*.8 = 829

It is correct. You are given WP. The second half of the WP has already been adjusted by the rate decrease. The first half needs to be adjusted by 20% decrease.

Hence, .5 * 933 * .8 + .5 * 933 = 840

2M
04-25-2006, 06:50 PM
It is correct. You are given WP. The second half of the WP has already been adjusted by the rate decrease. The first half needs to be adjusted by 20% decrease.

Hence, .5 * 933 * .8 + .5 * 933 = 840

Actually, I don't believe it is correct.

This, above gives the average written premium for the year. This doesn't on-level it


829 is the onlevel premium.

The Jones paper discusses comparing average premium @ current rate level for each year, in doing two-step trending.

(Mahler's solution also has 829 for the premium)

pcact
04-25-2006, 07:16 PM
Actually, I don't believe it is correct.

This, above gives the average written premium for the year. This doesn't on-level it


829 is the onlevel premium.

The Jones paper discusses comparing average premium @ current rate level for each year, in doing two-step trending.

(Mahler's solution also has 829 for the premium)

Thanks, I see the incorrect assumption in that solution. BTW, average written premium is given.

BassFreq
04-26-2006, 01:27 AM
Hello. My name is BassFreq and I'm a...

Oh! I'm sorry, I thought this was the 12-step thread.

2M
04-26-2006, 01:50 AM
average written premium is given.

"Average written premium" prior to the 20% rate change is given

pcact
04-26-2006, 08:40 AM
"Average written premium" prior to the 20% rate change is given

Given average written premiums: all of 2002 is prior to -20% rate change, half of 2003 is prior to -20% rate change, and all of 2004 is after -20% rate change.

2M
03-09-2007, 11:05 AM
:wave:

ReserveRage
04-07-2007, 08:28 PM
Wow, there's a lot of discussion on this topic. Just so that I have this clear.

In 2-step trending, in the first step, we always trend to 6 months prior to the last date in the experience period? This is because each date of the experience period is actually a 12 month average, so that if our experience period was 2000-2002, then 12/31/02 is really an average of written policies from (12/31/01 - 12/31/02) and this midpoint is 7/1/02?