Ref. Cook, p. 7 last sentence
Cook says, "This model will have twice as much variation (or 1.4 time =
as much
standard deviation) as the real world, due to random errors, and will h=
ave a
much shorter period for cyclical errors." Why "twice"? In general, do=
es this
sentence make sense to anyone?
Ref. Head, p. 4 says,
"The possibility of losses less than the policy face forces the insurer=
to make
an assumption about the amount of insurance purchased in order to set a=
premium
rate which equates the expected indemnity payments with pure premiums."=
Can
anyone explain this? Specifically, why would an insurer ever have to m=
ake an
assumption about the amount of insurance. Wouldn?t an insurer know the=
amount
of insurance? An insured takes out a policy for $100,000, his amount o=
f
insurance is $100,000. What?s to assume?
Ref. NEAS? Underwriting Profit provision #2, p. 19 (PVI/PVE)
Says, "WRT the denominator, a boost in the discount rate will always re=
duce the
PVE. This will enhance leverage so that positive PVI/PVE grows more po=
sitive
and negative PVI/PVE grows more negative."
Consider cash flows of -100 at end of year 1 and +200 at end of year 2
Further consider equity flow of +1000 at end of year 1
Assume i =3D 5% =3D> Then PVI/PVE =3D .0905
Assume i =3D 10% =3D> PVI/PVE =3D .0814
PVI/PVE was positive and grows LESS positive with the increase in disco=
unt rate.
What?d I do wrong? (Admittedly I don?t get any of the methods 4 throug=
h 7 -- at
all.)
Ref. Bluhm ? Experience Rating, p. 9 Figure 3
Figure 3 shows a histogram of actual losses divided by expected losses.=
It also
shows how the histogram would change after specific stop-loss. Assumin=
g the
catastrophe cut-off figure was actual/expected =3D 120%, I would have e=
xpected to
see the frequency at this point =3D the frequency of all claims greater=
than 120%
in the no-stop-loss curve. But it doesn?t. In the stop-loss curve; th=
e
frequency gradually decreases to 0 rather than abruptly stopping (e.g. =
at 120%).
Why?
Ref. Bluhm ? Experience Rating, p. 12
What is the characteristic of medical insurance that makes its "older e=
xperience
of limited use"?
Ref. Boor p. 326
How is "Tau-sub-1", the prediction error of the base statistic as a sta=
nd-alone
predictor of next year?s loss costs, computed? Just the sum of the (Ac=
tual ?
Expected)?
Ref. Head, p. 78
Says, ?Practically, statistics on the before-loss value of damaged prop=
erties
are not sufficient to permit the data to be stratified to obtain percen=
tage loss
severity figures for properties in different size classes.? I don?t ge=
t it ?
why wouldn?t the value of the property and the loss amount be sufficien=
t to
derive and percentage of loss distribution you could want?
=