what exactly is "loss cost?"

[leezard]

When practicing problems, the term "loss cost" always throws me.

From Clark, it seems that loss cost is reinsurer's expected losses as % of subject premium. See p. 15 of Clark paper.

Miccolis calls his g(x;k) function "the insurer's cost function." See p. 29 of Miccolis paper. This sounds like severity (net of deductibles and limits). See old exam 9 2006 #8.

So if I see a Clark problem, loss cost is aggregate losses, and in Miccolis problems it's severity? Any other interpretations I should know?

[rj_rattigan]

Originally Posted by leezard

When practicing problems, the term "loss cost" always throws me.

From Clark, it seems that loss cost is reinsurer's expected losses as % of subject premium. See p. 15 of Clark paper.

Miccolis calls his g(x;k) function "the insurer's cost function." See p. 29 of Miccolis paper. This sounds like severity (net of deductibles and limits). See old exam 9 2006 #8.

So if I see a Clark problem, loss cost is aggregate losses, and in Miccolis problems it's severity? Any other interpretations I should know?

Loss Cost = E[Loss] / Exposures = Pure Premium.

In reinsurance, exposures = subject premium.

I don't see any ambiguity calling something else the "insurer's cost function". Maybe I don't understand the question. The g() and h() functions go from ground up $ of loss -> insurer's share $ of loss. If you take the expectation and incorporate the expected number of claims per exposure you'd have the loss cost.

Hope this helps.

[leezard]

In Miccolis, pure prem is E[Y] = E[g(x)]*E[n], so g(x) must be severity.

In old exam 9 2006 #8, they give you a column called "Expected Loss Cost for the Occurrence Limit," then the solutions treat that column as g(x), and multiplies it by claim counts, so they're treating what they labeled "Loss Cost" as severities.

If you're confident "loss cost" means loss/exposure, and Miccolis never called his g(x) "loss cost" but rather "cost function," then my question turns into a complaint about that particular exam question using the terminology wrong (which wouldn't be a surprise).

[sloppyjoe]

Originally Posted by leezard

In Miccolis, pure prem is E[Y] = E[g(x)]*E[n], so g(x) must be severity.

In old exam 9 2006 #8, they give you a column called "Expected Loss Cost for the Occurrence Limit," then the solutions treat that column as g(x), and multiplies it by claim counts, so they're treating what they labeled "Loss Cost" as severities.

If you're confident "loss cost" means loss/exposure, and Miccolis never called his g(x) "loss cost" but rather "cost function," then my question turns into a complaint about that particular exam question using the terminology wrong (which wouldn't be a surprise).

In Miccolis:
X, g(x) = severity
Y = pure prem
N = frequency