e*******c
发帖数: 2133 | 1 目前在看health reserving的东西，有点疑问。首先，health reserving和p&C
reserving很像，但又有不少区别。比如p&C reserving里会有reported（incurred)
loss, paid loss的区别,还会要考虑p/i ratio.比如比较新的munich chain ladder就
是可以缩小p/i ratio的一个方法；但是，在health里好像没有reported loss这个概念
，只有paid loss,然后自己算completion factor(就是1/LDF,财险里的reported %（好
像是啊）),然后用paid loss/CF算出来estimated incurred loss,也就是说人家都用不
上munich chain ladder这么有技术含量的东西。
再有个小疑问就是健康险里算completion factor我觉得又不太像1/ldf,好像是1/(LDF1
*LDF2*LDF3....LDFn),关键是ldf这里已经是cumulative loss的了，再乘起来就挺奇怪
的了。举个数字的例子吧：1,2,4,8,8,8,8,8,(cumulative loss啊）ldf就是，2，2，2
，1。。。。然后cummulative ldf就是，8，4，2，1....,1/ldf就是12.5%,25%,50%,
100%....Health insurance 里也是12.5%,25%,50%,100%....但是算completion factor
好像是把这几个数再乘起来，得到：0.015625,0.125,0.5,1....我觉着挺奇怪的，这事。
麻烦版上几位了。@IBNR,@sword1900
Happy July 4th! |
s*******0
发帖数: 3461 | 2 ibnr版大是专家 我等着看回答啊
你的问题看的不是很明白啊 每一年的ratio算出来 最终的ratio当然需要全部乘起来
能不能在说的通俗点啊
ldf 应该是每一年的损失比率吧？ 就是说每一年比上年的损失比率的均值，最终的损失应该要吧因子乘起来
看你的描述应该是用的BF法则，先估计出最终的损失，然后再除以发展因子来估计未决赔款准备金吧。但是发展因子还是需要每年的乘起来，这个不会变的。 |
s*******0
发帖数: 3461 | 3 btw 原来今天独立日啊，怪不得很多餐馆不开，差点没饭吃！ |
e*******c
发帖数: 2133 | 4 你说的是age to age ldf，算cumulative肯定得都乘起来，然后取倒数应该就是report
%对吧？但是我看到的好像是取完倒数了，在cumulatively multiply一下，非常困惑
这事啊
损失应该要吧因子乘起来
决赔款准备金吧。但是发展因子还是需要每年的乘起来，这个不会变的。
【在 s*******0 的大作中提到】
: ibnr版大是专家 我等着看回答啊
: 你的问题看的不是很明白啊 每一年的ratio算出来 最终的ratio当然需要全部乘起来
: 能不能在说的通俗点啊
: ldf 应该是每一年的损失比率吧？ 就是说每一年比上年的损失比率的均值，最终的损失应该要吧因子乘起来
: 看你的描述应该是用的BF法则，先估计出最终的损失，然后再除以发展因子来估计未决赔款准备金吧。但是发展因子还是需要每年的乘起来，这个不会变的。
|
s*******0
发帖数: 3461 | 5 有没有模型 ？
把你msn 站内给我
如果不是机密的话 |
s*******0
发帖数: 3461 | |
G*****u
发帖数: 1222 | 7 p&c里面也不是全用reported losses
需要看line of business以及数据的可靠性
"关键是ldf这里已经是cumulative loss的了"?这句话怎么理解 没看懂
ldf全称是loss development factor
在triangle里面是通过当月的loss(either paid or incurred) 除以前月的loss
是一个ratio 怎么会是cumulative loss
关于你health insurance completion factor计算那段 很可能是你弄错了
我以前见过一些health reserving triangle
和P&c reserving triangle排列的朝向不太一样
但是最终计算结果应该是相同的 |
e*******c
发帖数: 2133 | 8 "关键是ldf这里已经是cumulative loss的了"我指的是age to ultimate LDF.ATU
factor不就是把所有的ldf乘一块吗？换句话说每一个age的cumulative loss/ultimate
loss=report %(CF in Health reserving),然后1/report %=ATU LDF.
不知您能不能先讲讲incurred loss和paid loss的区别和p/i ratio的事
【在 G*****u 的大作中提到】
: p&c里面也不是全用reported losses
: 需要看line of business以及数据的可靠性
: "关键是ldf这里已经是cumulative loss的了"?这句话怎么理解 没看懂
: ldf全称是loss development factor
: 在triangle里面是通过当月的loss(either paid or incurred) 除以前月的loss
: 是一个ratio 怎么会是cumulative loss
: 关于你health insurance completion factor计算那段 很可能是你弄错了
: 我以前见过一些health reserving triangle
: 和P&c reserving triangle排列的朝向不太一样
: 但是最终计算结果应该是相同的
|
b********6
发帖数: 195 | 9 Incurred loss=paid loss +case reserve
p/i ratio=paid loss/incurred loss |
e*******c
发帖数: 2133 | 10 在health 里也一样，reserve=incurred loss-paid loss,但是incurred loss=paid
loss*（ATU LDF),这样的话incurred loss其实是取决于paid loss的，我不知道在p&c
里是否也是一样
【在 b********6 的大作中提到】
: Incurred loss=paid loss +case reserve
: p/i ratio=paid loss/incurred loss
|
|
|
I**R
发帖数: 1309 | 11 sorry but I don't know much about health insurance. what
you described about the p&c method seems right to me.
LDF1
，2
【在 e*******c 的大作中提到】
: 目前在看health reserving的东西，有点疑问。首先，health reserving和p&C
: reserving很像，但又有不少区别。比如p&C reserving里会有reported（incurred)
: loss, paid loss的区别,还会要考虑p/i ratio.比如比较新的munich chain ladder就
: 是可以缩小p/i ratio的一个方法；但是，在health里好像没有reported loss这个概念
: ，只有paid loss,然后自己算completion factor(就是1/LDF,财险里的reported %（好
: 像是啊）),然后用paid loss/CF算出来estimated incurred loss,也就是说人家都用不
: 上munich chain ladder这么有技术含量的东西。
: 再有个小疑问就是健康险里算completion factor我觉得又不太像1/ldf,好像是1/(LDF1
: *LDF2*LDF3....LDFn),关键是ldf这里已经是cumulative loss的了，再乘起来就挺奇怪
: 的了。举个数字的例子吧：1,2,4,8,8,8,8,8,(cumulative loss啊）ldf就是，2，2，2
|
e*******c
发帖数: 2133 | 12 Question: Is reported loss totally dependent on paid loss?If so, then P/I
ratio is just 1/(ATU LDF), since incurred loss=paid loss*(ATU LDF)
【在 I**R 的大作中提到】
: sorry but I don't know much about health insurance. what
: you described about the p&c method seems right to me.
:
: LDF1
: ，2
|
I**R
发帖数: 1309 | 13
c
ATU means age-to-ultimate? if so , I think your "incurred loss" is different
from the "incurred loss" we normaly quoted. the post above has the most
commonly used definition.
your "incurred loss" sounds more like ultimate loss calculated using
chain ladder method(paid)
【在 e*******c 的大作中提到】
: 在health 里也一样，reserve=incurred loss-paid loss,但是incurred loss=paid
: loss*（ATU LDF),这样的话incurred loss其实是取决于paid loss的，我不知道在p&c
: 里是否也是一样
|
I**R
发帖数: 1309 | 14 again, as blabla0706 stated above, in p&c
incurred = paid + case reserve
【在 e*******c 的大作中提到】
: Question: Is reported loss totally dependent on paid loss?If so, then P/I
: ratio is just 1/(ATU LDF), since incurred loss=paid loss*(ATU LDF)
|
e*******c
发帖数: 2133 | 15 how do you get incurred loss? It's reported from client or it's calculated
based on paid loss?
【在 I**R 的大作中提到】
: again, as blabla0706 stated above, in p&c
: incurred = paid + case reserve
|
I**R
发帖数: 1309 | 16 incurred = paid + case reserve
when a claim is reported to the insurer, a claim number is assigned
to it, and depending on the nature of the claim, claim department
would set up a case reserve for that. at any time, the cumulative
paid amount + the case reserve = incurred.
on day 1, probably your paid is 0(since you rarely pay anything
right away), and your case is some non-zero value. as you have
more info regarding the claim, you would pay more and more(assuming
no negative paid), and your case reserve normally decrease(assuming
your initial estimate of the claim amount is fairly accurate).
after you have settled the claim, you close the claim by reducing
the case reserve to 0. now your paid = incurred, and this is also
you ultimate claim value. reserving is to estimate this value
periodically.
hope this helps.
【在 e*******c 的大作中提到】
: how do you get incurred loss? It's reported from client or it's calculated
: based on paid loss?
|
a********t
发帖数: 1810 | 17 incurred loss in health is not equal to incurred loss in P&C
incurred loss=paid loss/completion factor in health
incurred loss in P&C is more complicated |
e*******c
发帖数: 2133 | 18 Thanks, that makes a lot of sense. Case reserve=reported loss-paid loss is
always true whether in P&C or Health, right?
I was thinking about using munich chain ladder method in health reserving
and found the different definition of incurred loss between P&C and Health
【在 a********t 的大作中提到】
: incurred loss in health is not equal to incurred loss in P&C
: incurred loss=paid loss/completion factor in health
: incurred loss in P&C is more complicated
|
a********t
发帖数: 1810 | 19 yes, it is true, but reported loss is only used in P&C, I don't think there
is name item "reported loss" in health |
e*******c
发帖数: 2133 | 20 right, incurred loss in health is actually "estimated incurred loss". In p&c
, incurred loss=reported loss
there
【在 a********t 的大作中提到】
: yes, it is true, but reported loss is only used in P&C, I don't think there
: is name item "reported loss" in health
|
|
|
a********t
发帖数: 1810 | 21 对吗?P&C里,有VALUATION DATE,在那一天REPORTED的LOSS叫REPORTED LOSS,PAY出去
LOSS叫PAID LOSS,还有一个ULTIMATE LOSS(INCURRED LOSS?),它和REPORTED LOSS的差
叫IBNR吧?
在HEALTH里,是先求INCURRED LOSS,用IL求COMPLETION FACTOR,你说反了. |
e*******c
发帖数: 2133 | 22 没，incurred loss=reported loss,在p&c里，ultimate loss是ultimate loss，单另
儿的
【在 a********t 的大作中提到】
: 对吗?P&C里,有VALUATION DATE,在那一天REPORTED的LOSS叫REPORTED LOSS,PAY出去
: LOSS叫PAID LOSS,还有一个ULTIMATE LOSS(INCURRED LOSS?),它和REPORTED LOSS的差
: 叫IBNR吧?
: 在HEALTH里,是先求INCURRED LOSS,用IL求COMPLETION FACTOR,你说反了.
|
g****o
发帖数: 124 | 23 In health, we don't say "reported loss". total incurred = paid + IBNR (
incurred but not reported) + reported but not paid. some actuary use
incurred = paid + IBNA (incurred by not adjudicated) + adjudicated but not
paid. Typically, you don't need a reserve for the "reported but not paid" as
the number should be known or close to be known. It is the IBNR or IBNA
that needs a reserve. We do use lag triangles and CFs to estimate INBR and I
agree with one of the posts that the method is same as what's used in P&C
except the rows and columns are switched. |
I**R
发帖数: 1309 | 24 thanks for the clarificaiton
as
I
【在 g****o 的大作中提到】
: In health, we don't say "reported loss". total incurred = paid + IBNR (
: incurred but not reported) + reported but not paid. some actuary use
: incurred = paid + IBNA (incurred by not adjudicated) + adjudicated but not
: paid. Typically, you don't need a reserve for the "reported but not paid" as
: the number should be known or close to be known. It is the IBNR or IBNA
: that needs a reserve. We do use lag triangles and CFs to estimate INBR and I
: agree with one of the posts that the method is same as what's used in P&C
: except the rows and columns are switched.
|
s*******0
发帖数: 3461 | 25 rnbr ( reported but not recorded)
有么 ？
【在 I**R 的大作中提到】
: thanks for the clarificaiton
:
: as
: I
|
s*******0
发帖数: 3461 | 26 顶一下 这边是否会有 rbnr? thanks |
g****o
发帖数: 124 | 27 Once a claim is filed, it should be recorded in the claims system right away
. However, depending on how complicated the claim is, it may some time to
adjudicate the claim, i.e. Determine whether the claim is eligible for
payout, how much the plan should pay, etc.. It also takes time to actually
cut the check and send it out. |
