When Numbers Start Making Decisions · Season Four, “Things That Cannot Be Priced Directly” · Article 1
1. What if a safety rule is expected to save 0.4 of a person?
Imagine a proposed rule requiring additional guards on industrial machinery. Over nine years it is expected to prevent about 0.4 fatal accidents, while industry would spend A$1.5 million complying. No identifiable worker called “0.4 of a person” is waiting to be rescued. Nobody knows who would otherwise die.
Saying that life is priceless does not tell government whether the rule is proportionate. Valuing only lost earnings would imply that children, retirees and unpaid carers are worth less. Australia’s Office of Impact Analysis therefore supplies a common estimate for cost-benefit analysis. Its February 2026 guidance recommends a Value of Statistical Life, or VSL, of A$5.87 million in 2025 prices and a Value of a Statistical Life Year of A$253,000. Office of Impact Analysis: Value of statistical life
At that scale, a reduction of 0.4 expected deaths is valued at about A$2.348 million before other costs, benefits and uncertainty are considered. The calculation does not value an identified person. It aggregates willingness to pay for many small reductions in mortality risk. If 100,000 people each experience a one-in-100,000 reduction, the expected total is one statistical life.
The real question is how much decision-making authority this manufactured number should receive.
2. A$5.87 million does not come from a catalogue of lives
VSL research usually studies willingness to pay for risk reduction. Stated-preference studies ask what people would pay for safer conditions. Revealed-preference studies observe choices such as purchases of vehicle safety features or wage differences associated with occupational risk.
Both methods have limitations. Survey answers may not reflect real budgets. Wage differences mix risk with bargaining power, industry structure and imperfect information. A low-income person may be willing to pay less because of capacity, not because survival matters less. Market observations can therefore reproduce existing inequality.
The OIA uses a standard estimate because commissioning a complete study for every proposal would be costly and inconsistent. Updating the estimate with the Wage Price Index maintains a common policy language. Standardisation allows safety proposals in different sectors to be compared rather than defended through selectively chosen values.
But a standard VSL should not acquire an ontological meaning. It is an estimate of the social value of small, unforeseen risk reductions in policy analysis. It is not the price of a person, a compensation schedule or a ceiling on treatment.
3. Why institutions still need this unsettling number
Government already values risk, whether it admits it or not. Every decision about road design, workplace regulation, medicine safety and emergency preparedness allocates resources among hazards. Refusing to quantify can hide the trade-off inside political instinct or departmental habit.
VSL makes one dimension explicit. Analysts can compare expected mortality benefits with compliance costs and with alternative uses of resources. It also prevents salary from being the main measure of loss and gives a common value to risk reduction across people.
The Office of Impact Analysis nevertheless warns against inappropriate use. VSL is designed for small changes in the risk of unforeseen fatality. It should not be used as an automatic value for identified lives, voluntary risk or every health outcome. Injury, illness and disability may require different evidence.
A calculation is therefore strongest as structured evidence in a decision, not the decision itself. Cost-benefit analysis should show who is affected, how risk was estimated, which values are unquantified and how uncertainty changes the result. OIA: Cost Benefit Analysis
4. Equal totals can conceal radically different distributions
Two policies may each reduce one statistical death yet distribute risk differently. One gives a tiny benefit to millions of drivers. Another removes a severe hazard faced by a small group of workers. A simple expected total can treat them as equivalent.
Baseline risk, voluntariness, control and vulnerability matter. A person compelled to face an occupational hazard is not in the same position as a consumer choosing among products. A remote community exposed to a single industrial facility may carry concentrated risk and limited political power. Children cannot bargain over safety.
Distribution can also determine confidence. A model built from population averages may not describe a high-risk subgroup. If the people bearing the danger differ from those whose preferences informed the VSL estimate, the transfer requires explanation.
Cost-benefit analysis should therefore report more than net present value. It should identify winners and losers, risk concentration, uncertainty and groups with limited capacity to avoid exposure. A positive aggregate result cannot by itself authorise an unfair distribution.
5. The most dangerous error is changing the question
VSL answers a policy question: what monetary value can be assigned to a small expected reduction in fatality risk across a population? It does not answer how much should be spent rescuing an identified person, what damages a family deserves after wrongful death, or whether one person’s death can be offset by unrelated profit.
Once a figure exists, administrative reuse is tempting. A department may divide a program cost by lives saved and treat the VSL as a strict approval line. A hospital may appear to face the same problem. A court may seek a convenient damages figure. These uses silently replace the object being valued.
Identified rescue involves duties, rights, urgency and clinical evidence that are absent from ex ante risk regulation. Compensation addresses responsibility and loss after harm. The VSL is not validated for either purpose. Accuracy within one model does not create universal authority.
Every use should name the object: risk reduction, life-years, compensation, budget impact or treatment. If the noun changes, the method must be justified again.
6. Price is an interface for action, not the essence of life
Life is lived through relationships, capacities, memories and obligations. None is contained in A$5.87 million. Policy nevertheless needs an interface through which dispersed risks can enter a common decision. VSL is that interface.
The distinction avoids two extremes. It rejects the claim that life is literally worth a market sum. It also rejects the fantasy that a government can allocate safety resources without making comparative judgements. Quantification is a practical structure placed between evidence and action.
Its legitimacy depends on keeping the structure open. Ethical and distributional constraints must remain able to overrule an attractive total. Uncertainty must be visible. The number must be updated and its scope explained. Decision-makers must remain responsible for the policy rather than saying that the calculation chose.
7. Minimum safeguards for pricing risk reduction
Before VSL influences a decision, ask:
- Is the proposal about small ex ante risk reductions rather than an identified life?
- How was the baseline risk and expected change estimated?
- Which population faces the risk, and does the evidence represent it?
- Is exposure voluntary, avoidable or imposed?
- How are benefits and burdens distributed?
- Which health, cultural or rights-based effects remain outside the monetary total?
- Does sensitivity analysis change the recommendation?
- Who explains and accepts responsibility for the final judgement?
These safeguards allow a common number to clarify trade-offs without granting it power to define human worth.
Conclusion: price the reduction in risk, not the person
The Value of Statistical Life makes safety benefits visible in decisions that would otherwise favour easily counted costs. A common estimate can improve consistency and prevent earnings from determining whose risk matters.
My judgement is that A$5.87 million is legitimate as an updated policy value for small reductions in unforeseen fatality risk, provided its population, uncertainty and distribution are shown. It must never become a price attached to a particular life, a compensation ceiling or an automatic rescue limit.
Season Four begins with a boundary: institutions may monetise a dimension needed for comparison without claiming that everything valuable has entered the price.
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.