File Image: IOL

For people buying life insurance, discrimination based on their level of risk can be ethically justified. However, on one of the underwriting criteria, socio-economic class, South African insurers may be open to the accusation of unfair discrimination on the way they allocate applicants to the different premium levels.

This is the conclusion of retired actuary Francois Marais, in his thesis for the degree of Master of Philosophy in Applied Ethics, “A critical evaluation of discrimination in risk underwriting in the life insurance industry in South Africa”, which he presented at the recent Colloquium 2019 of the International Actuarial Association in Cape Town.

Marais bases his argument on the ethical theory of “moral contractualism”, advanced by the American philosopher, Thomas M Scanlon. In his influential work What We Owe to Each Other (1998), Scanlon states that an act may be regarded as wrong “if it is disallowed by any principle for the general regulation of behaviour that no one could reasonably reject.”

Marais writes: “Scanlon’s theory … provides an appealing lens for considering the reasons and the justification of premium discrimination in life insurance. At the core of Scanlon’s theory is his definition of moral wrongness, based on the concepts of reason and justifiability. Unlike utilitarianism, with its one central moral value of well-being, moral contractualism can accommodate a plurality of ethical notions within its unified normative domain of justifiability.”

The four factors that South African insurers use to determine what risks you pose – and thus what size premium you pay – on a retail life policy are: age, sex, smoking status, and socio-economic class (see table). Standard mortality rates for the four factors are published regularly by the Actuarial Society of South Africa and are used across the industry.

Apart from from the normal premium based on the four general factors, you may also be charged an individual premium loading, based on your medical history, occupation, and leisure pursuits.

Marais says that the differences in premiums depending on where you fall within each of the four general groups can be considerable. A young non-smoking professional woman will get about R1 million life cover for a premium of R100 a month. An old, uneducated, poorly-paid male smoker will get only R22 000 cover for the same premium. In other words, people at the extreme low-risk end of the underwriting spectrum will receive almost 50 times more cover than people at the extreme high-risk end.

In determining whether this discrimination is justified, Marais defines two contractualist principles that apply to underwriting, which he argues that nobody could reasonably reject. He then applies these principles to the underwriting practices, to see whether they can be ethically justified.


1. The Fair Lottery Principle

In a lottery, where each ticket has an equal chance to win, all tickets should cost the same. It would be unfair to charge different prices for the tickets.


In a fair lottery, your chances of winning are determined by how many tickets you buy, and thus how much you pay in total. If you bought 10 tickets, your chances of winning would be 10 times as much than that of someone buying a single ticket. While the person who bought only one ticket may still walk off with the winning prize, he presents one tenth of the risk to the lottery company than you do with your 10 tickets. Marais argues that this principle would pass Scanlon’s test that it cannot be reasonably rejected.

If the principle, is applied to insurance it shows that the premium you pay should be proportionate to the probability of you receiving a payout. “We can justify the general principle of discrimination in life insurance underwriting,” Marais argues.

The second issue is: can insurers justify each of the four underwriting factors?


2. The Fair Discrimination Principle

An underwriting factor is justifiable if it has:

  • Strong and reliable statistical evidence;
  • Reasonable causal explanation; and
  • Unambiguous allocation to risk groups.


“Actuarial risk assessment is not an exact science,” Marais says. “We can show strong evidence of significant mortality differences for each factor, but cannot accurately quantify risk related to each factor.”

He says there is strong, reliable statistical evidence, based on the large amount of data collected by insurers over decades, of significant mortality differences under each of the four underwriting factors. Apart from the obvious factor of age, the mortality of men is about 50% higher than women; for smokers, mortality is 50% higher than non-smokers for smokers under 45 years of age and 80% higher for smokers over 45; and the mortality of people in the lowest socio-economic class is 300% higher than that of people in the highest class (see table).

On the test of reasonable causal explanation, Marais reminds us that statistical correlation is not proof of cause. However, there are reasonable explanations for the differences.

  • Sex: Not only do women live longer than men, but their genetic advantage is clear from birth, Marais says. In addition, social factors such as working conditions, dangerous activities, road accidents and alcohol use tend to increase male mortality.
  • Smoking: There is overwhelming medical evidence of its adverse impact on health.
  • Socio-economic status: People with higher education and income levels generally have healthier lifestyles leading to lower mortality. They tend to have healthier diets, better medical care, exercise more, live in safer environments, have less hazardous jobs and consume less alcohol.

On the third test, the allocation of applicants to risk groups, for three of the four factors, the classification is discrete and unambiguous: you are of a specific age, either male or female, and either a smoker or non-smoker (based on your declaration on your application). However, it is on the allocation of people into classes based on socio-economic status that problems arise:

  • There is no common objective standard of socio-economic class across the industry. Each insurer has its own methodology for allocating applicants to one of four or five classes according to a combination of education level and income band. Income bands are regularly adjusted for inflation by each company.
  • While education level is relatively clear-cut and stable, income level is more problematic. Specifically, the broad income bands used are fairly arbitrary and not easy to justify. For example, if the cut-off level between classes is R16 000 a month (as in the example in the table), should someone earning R16 001 a month be paying a significantly lower premium than someone earning R15 999 a month?

It is on the allocation of applicants to socio-economic class that the industry fails the Fair Discrimination Principle defined in terms of Scanlon’s theory of moral contractualism and may therefore be exposing itself to the criticism of unfair discrimination, Marais suggests.

He says the problem of these “large discrete jumps in premiums at rather arbitrary salary cut-off points” could be remedied relatively simply “by using a graded salary scale rather than just four or five classes”.


DEFINITIONS:

Mortality: the death statistics of a specific population group.

Underwriting: the assessment of risk (in this context).