Feb 26, 2013

NOT Discriminating is NOT possible

Tomorrow I'll be discussing the borders of solidarity as a panel member at an actuarial congress (VSAE)  for econometricians in The Hague (The Netherlands).

In a Dutch interview preceding the congress, two students asked me:

"The  Court of Justice of the European Union (CJEU) has decided that the use of gender as a risk factor by insurers should not lead to individual differences in premiums and benefits.
What is your opinion?"

My short answer was :

NOT Discriminating is 
NOT Possible

Let me illustrate this 'quantum quote' with two examples.

Example I: Gender Neutral Car Insurance 
  • It's scientifically proven that women are better drivers, have just as much car accidents as men, but cause less damage. That's a fact and that's why car insurance for women is cheaper than for men. 
  • As from December 21, 2012, European insurers are not allowed to 'discriminate' anymore by gender, implying equal car insurance premiums for men and women. 
  • If insurers calculate this premium as the weighted average of their portfolio, women are obliged to pay (much) more premium than before and also more than actually and actuarially necessary regarding their gender group. 
  • Therefore women are de facto discriminated, although the genuine intention was NOT to discriminate!
  • Not only women, but also insurers are discriminated as they now will be faced with anti-selection: Relatively more men will choose an insurance cover, as car insurance premiums for men have become less than the premium corresponding with the expected damage for their gender group. Insurers will therefore face a loss on car insurance. 
  • Based on solvency legislation, the insurer will (next) be 'forced' to increase the average weighted premium. This - in turn - is at odds with the measures envisaged by the European Court. 
  • A similar kind of reasoning applies also for unisex rates for pension and life insurance.
  • The upcoming (2014) US health care law will also prohibit “gender rating”. However, gaps persist in most states. There seem to be no signs of insurers that have taken steps to reduce them.

The conclusion must be that discrimination regulation is carried too far.

 'Over-Solidarity' as in this case has nothing to do with real solidarity and is in nobody's interest; it has become 'Anti-solidarity'.

The proposed measures - no matter how well intended - have a opposite effect and should be reconsidered on basis of the question: are the discriminating effects before the new legislation more or less than after?

We've got to stop discrimination due to over-discrimination and anti-discrimination!

Insurance Rating Fallacy: Gender anti-discrimination laws are superseded 
Prohibiting "gender", "marital status" and "age" as rating elements doesn't solve anything.

Modern rating systems based on data mining (Google history), social media (premium quoting on basis of: your smart-phone that captures and shares your drivingstyle with the insurer) and neural networks are "black boxes" that quote insurance premiums in such a way that every client can get individually quoted on bases of his 'profile'.

That 'profile' doesn't have to contain any of the forbidden discriminating elements (nor direct related) to get satisfactory results for clients as well as insurers. Although there are also simple (e.g. Bayesian-Classification) techniques to derive a clients gender from other non-discriminant related variables (e.g. height, weight and foot-size determine gender quite accurately) in an insurers direct or indirect related data base, insurers and their actuaries would end up in an unwanted ethical dilemma by using these direct-related techniques.

Another illustrative and strong example of determining your gender on bases of - at first sight - non-gender-related information is Hacker Factor's "Gender Guesser"  that attempts to determine an author's gender based on the words used. Try  "Gender Guesser" for yourself HERE. Take a part of an email you've written (more than 300 words), copy-paste it to Gender Guesser and notice how gender Guesser will probably determine your gender without any problem in a split of a second!

These simple techniques show that the developed anti discrimination legislation is superseded and has become irrelevant for insurers and their clients to come anyhow to an adequate and ethical responsible rating policy on basis of neural networks or social media related information, such as information from smartphones that transmit your driving style information to the insurer (why not, if you have nothing to hide?).

Example 2:.Women on Boards: Commission proposes 40% objective
The European Commission has proposed legislation with the aim of attaining a 40% women presence objective in non-executive board-member positions of publicly listed large companies.
Currently, large boards are dominated by men (85% non-executive, 91% executive).

No matter how welcome and needful women are on board level, forcing such a development makes no sense and will have an adverse effect.

From experience I can tell that women who really qualify for board level positions, are very unhappy if they are appointed under the vigor of gender legislation and not on basis of their acknowledged competences.

This is perhaps a sign that women who really qualify feel discriminated by this new proposal. Proposals should better emphasize on stimulating women presence on board level and take away old boys network principles.

Anti discriminating legislation often results in the exact opposite of what is intended. Legislation is often superseded, should be carefully evaluated on its effects and certainly reconsidered if the discriminating effects after applying the new legislation increase.

Used Sources and Links

Humor: Actuarial Creativity

As actuaries we've studied a lot in life. And to keep up with actuarial science we'll probably keep studying until our personal mortality rate hits us finally in the back.

Although study brought us to the top of financial and statistic modeling, there's a small but fatal risk that we become so engrossed in our work that we loose our creativity or ability to solve things in a simple way.


To test whether you're still a creative 'simplist', let's do a short 3 question test. Here it is:

Question 1
 "Show how it is possible to determine the height of a tall building with the aid of a barometer."

If you think you've solved this high school level problem, go to the next question

Question 2
 "Solve question 1 with another method."

If you think you've solved this problem as wel, go to the final question

Question 3
 "Solve question 1 with 4 other methods."

Although actuaries never give up, there's a slight chance you had to surrender and are longing for the answer.
In that case (only), read further for the answer.

Answer: The Barometer Fable
Bob Pease (Nat.Semi.) records the story of the Physics student who got the following question in an exam: "You are given an accurate barometer, how would you use it to determine the height of a skyscraper ?"

  1. He answered: "Go to the top floor, tie a long piece of string to the barometer, let it down 'till it touches the ground and measure the length of the string".

    The examiner wasn't satisfied, so they decided to interview the guy: "Can you give us another method, one which demonstrates your knowledge of Physics ?"
  2.  "Sure, go to the top floor, drop the barometer off, and measure how long before it hits the ground……"

    "Not, quite what we wanted, care to try again ?"
  3. "Make a pendulum of the barometer, measure its period at the bottom, then measure its period at the top……"

    "..another try ?…."
  4. "Measure the length of the barometer, then mount it vertically on the ground on a sunny day and measure its shadow, measure the shadow of the skyscraper….."

    "….and again ?…."
  5. "walk up the stairs and use the barometer as a ruler to measure the height of the walls in the stairwells."

    "…One more try ?"
  6. "Find where the janitor lives, knock on his door and say
    'Please, Mr Janitor, if I give you this nice Barometer, will you tell me the height of this building ?"

Find more than 140 solutions and read the original famous Barometer Fable, as published in 1968 in an article  by Alexander Calandra.

Keep in mind that not every method leads to satisfactory results.
An uncertainty analysis of determining a building height using a barometer, developed by Israel Urieli, shows that this method is not accurate at all!

So the surprising news is that the first two alternative methods mentioned above are more accurate than the method you learn at high school.

It's always best when you can solve an (actuarial) problem in more than one way and the outcomes point in the same direction. The more a specific solution comes to front by applying different methods and/or data, the more confident you can be that the outcome is robust.

Used Sources

Feb 5, 2013

Supervision on Supervision

On February 1 2013, the Dutch Minister of Finance, in close consultation with the Dutch Supervisor 'De Nederlandsche Bank' (DNB), announced he nationalized the Dutch Bank-Insurer SNS Reaal.

Intervention was necessary to prevent grave threats to the Dutch financial stability and economy.

This intervention shows again that the European stress tests fail, as was already predicted in a Quartz article called "Forget the stress tests: Europe’s banks are a worrisome mystery" on October 2, 2012. Risk managers have to to a better job. Work to be done!

Role of the Supervisor
The intervention also raised the question about the role of the Dutch supervisor DNB in this debacle. Officially the (Dutch) Minister of Finance is responsible for the supervision on the national supervisor. In practice this role is delegated to the national 'Supreme Audit Institutions' (SAIs).

A special  European Committee Working  Group assessed the scope of the mandate of Supreme Audit Institutions (SAIs) and its proper functioning with respect to the main financial supervisor  (FSA)  for
prudential oversight on banks.

Thirteen (of the twenty seven) European countries participated in the SAIs research.

Three aspects were analyzed:
  1. Mandate: Has the local SAI a mandate to audit the supervisory role?
  2. Access:  Has the local SAI actually access to audit bank files of the supervisor\supervisor
  3. Test: Did SAI successfully test the completeness of the bank files

Here are  that are disappointing results of the work group for the main 11 countries:

Yes* = Yes , with condition of confidentiality

Although a general approach of (SAI) supervision on (Supervisor) seems useless and even silly, it's clear that the current supervisory grip and transparency is undeniably inadequate.

In this case, we certainly need a strong supervision on national supervisors in Europe to prevent accidents like SNS. In other words: Back to the old 'Four Eyes Principle'...

In a letter to the Dutch  House of Representatives the Dutch SAI states:
"The Council of Ministers agreed to the introduction of a European supervisory mechanism for banks, with a central role for the ECB, on 13 December 2012. 

To safeguard the information position of the European parliament and the member states, the European Court of Auditors should be able to audit the supervision exercised by the ECB.

The European Court of Auditors' current mandate does not allow it to do so. This creates an audit gap at European level: arrangements are not in place for the independent audit of the ECB's organisation and exercise of its 
supervisory tasks and authority. "

- State of the Netherlands nationalises SNS REAAL
- Forget the stress tests: Europe’s banks are a worrisome mystery
Points for consideration in the Dutch House of Representatives
- Points for consideration in the Dutch House of Representatives (Dutch)
- 4 Eyes Principle Cartoon