Two ways of using the Central Limit Theorem on a basic example

36 Views Asked by At

The problem is as follows.

An insurance company sells 10,000 similar car insurance policies. They estimate that the amount paid out in claims on a typical policy has mean £240 and standard deviation £800. Estimate how much they need to put aside in reserves to be 99% sure that the reserve will exceed the total amount claimed.

Let $S_n$ be the total amount claimed. The solution given is:

$$ P\left(\frac{S_n - nμ}{σ\sqrt{n}}< Φ^{-1}(0.99)\right) = 0.99. $$

Then you rearrange the inequality for $S_n$ and evaluate the RHS. I do not understand why there is $Φ^{-1}(0.99)$ involved. Below is my thought process.

Let $y$ be the amount in reserves. I want to find $y$ such that $P(y - S_n > 0) = 0.99$. By CLT, the random variable $X_n := \frac{S_n - nμ}{σ\sqrt{n}} \sim N(0,1) $ for large $n$. I can write $y - S_n$ in terms of $X_n$ and then it is:

$$ P( y - (σ\sqrt n X_n + nμ)) > 0 ) = P\left(X_n < \frac{y - nμ}{σ\sqrt n} \right) = 0.99 $$

Then using the fact that for the standard normal distribution $P(X < 2.326) \approx 0.99$ allows me to solve for $y$. Both of these give the same answer but I would like to know what is the thought process for getting the first expression?

1

There are 1 best solutions below

1
On BEST ANSWER

Let $R$ be the reserves, which have to be put aside and $S_n$ be the the total amount of claims. We need to find $R$, such that $S_n$, doesn‘t exceed it with probability $0.99$, so:

$$P(S_n<R)=P(\frac{S_n-n\mu}{\sqrt{n}\sigma}<\frac{R-n\mu}{\sqrt{n}\sigma})\sim \Phi(\frac{R-n\mu}{\sqrt{n}\sigma})$$

Furthermore: $$\Phi(\frac{R-n\mu}{\sqrt{n}\sigma})=0.99$$ which is equivalent to: $$\frac{R-n\mu}{\sqrt{n}\sigma}=\Phi^{-1}(0.99)$$ So we see: $$P(S_n<R) \sim P(\frac{S_n-n\mu}{\sqrt{n}\sigma}<\Phi^{-1}(0.99))=0.99$$