What's the right understanding of the concept of 'confidence level' for a financial institution?
That's not an easy question....
A short (popular) definition of confidence level in terms of Solvency and Basel regulation would be:
The probability that a financial institution doesn't default within a year.
In this blog I'll discuss and compare three more or less accepted confidence levels (CFLs):
Understanding Confidence Level
Before we get into the details, let's first shine a light on a widespread misunderstanding regarding the concept of 'confidence level'.
To make the concept of confidence level more understandable, one might argue as follows:
This method of reasoning is completely
The mistake that's been made is more or less the same as the next two fallacies:
Explanation
The probability of a pension fund with a confidence level of 97,5% going default, can be approximated by a simple Poisson distribution as follows:
From this we can conclude:
Insurer Confidence Level
For an insurance company with a confidence level of 99.5% the results are:
So even an insurer has a 4.88% default probability in a 10 years period on basis of a 99.5% confidence level. Keep this in mind if you take out a life insurance policy!!!
Banking Confidence Level
It starts getting serious when it comes down to a 99,9% confidence level for banks:
Comparison
Comparing the default probability of (Dutch) pension funds, insurers and bank on the long run:
Finally
Although this blog gives some more insight about the consequences of confidence levels on the long run, the real question of course is: what's the price you have to pay to avoid default risks?
That's something for another blog.....
Sources/Links
- Spreadsheet with tables used in this blog
That's not an easy question....
A short (popular) definition of confidence level in terms of Solvency and Basel regulation would be:
The probability that a financial institution doesn't default within a year.
In this blog I'll discuss and compare three more or less accepted confidence levels (CFLs):
- Dutch Pension Funds: CFL= 97.5%
- Life Insurers (Solvency II): CFL = 99,5%
- Banks (Basel II/III): CFL = 99.9%
Understanding Confidence Level
Before we get into the details, let's first shine a light on a widespread misunderstanding regarding the concept of 'confidence level'.
To make the concept of confidence level more understandable, one might argue as follows:
- The confidence level of a Dutch pension fund is defined as 97.5%
- This implies that there's a one years probability that the pension fund has an one year default probability of 2.5% (= 100% - 97.5%)
- This implies that the pension fund on average defaults once every 40 years (= 1 / 0.025)
This method of reasoning is completely
WRONG
The mistake that's been made is more or less the same as the next two fallacies:
- If one ship crosses the ocean in 12 days. 12 ships will cross the ocean in one day
- I fit in my jacket, my jacket fits in my suitcase, therefore I fit in y suitcase
Explanation
The probability of a pension fund with a confidence level of 97,5% going default, can be approximated by a simple Poisson distribution as follows:
- In 40 years the pension fund has a 63% default probability.
- The probability that the pension fund defaults more than once is 26%
- The probability that the pension fund defaults exactly once in a 10 years period is 19.47%
For an insurance company with a confidence level of 99.5% the results are:
So even an insurer has a 4.88% default probability in a 10 years period on basis of a 99.5% confidence level. Keep this in mind if you take out a life insurance policy!!!
Banking Confidence Level
It starts getting serious when it comes down to a 99,9% confidence level for banks:
Comparison
Comparing the default probability of (Dutch) pension funds, insurers and bank on the long run:
Finally
Although this blog gives some more insight about the consequences of confidence levels on the long run, the real question of course is: what's the price you have to pay to avoid default risks?
That's something for another blog.....
Sources/Links
- Spreadsheet with tables used in this blog