Benefit Transfer

Benefit Transfer of Willingness-to-Pay and VSL Values into a New Health Appraisal

Monetised health benefits in a cost-benefit analysis are typically not measured for that analysis. They are taken from earlier studies of willingness to pay, often estimates of the value of statistical life, and carried to a new population, country or policy with some adjustment, a practice known as benefit transfer. How that is done can turn a programme's net benefit from negative to positive, so it deserves the same scrutiny as the original valuation.

Study site, policy site and the value being moved

Every transfer links two settings. The study site is where the original valuation was done, through contingent valuation, a discrete choice experiment or a wage-risk study. The policy site is the population and decision to which the value is applied. Johnston and colleagues describe the practice as using pre-existing empirical estimates from settings already studied to predict economic values for other settings, a feasible route when time, funding and other constraints rule out an original study.

In health appraisal the quantity moved is usually a willingness to pay for a defined change, such as a small reduction in mortality risk or an avoided episode of illness. Robinson, Hammitt and O'Keeffe note that the practice is also called value transfer, and that few direct estimates of mortality risk values exist for low- and middle-income countries, where analysts typically extrapolate from wealthier countries. It differs from the transferability of a whole cost-effectiveness result between countries, which also raises questions of clinical practice, unit costs and comparators.

Unit, adjusted unit and function transfers

Transfers differ in how much study-site information they carry. The simplest moves a single number; the most elaborate moves an estimated relationship between value and its determinants.

The US Office of Management and Budget's Circular A-4 prefers transferring the whole function to a point estimate where there is a choice. The evidence is less one-sided. In a controlled five-country experiment on water quality, Bateman and colleagues found that simple mean value transfers performed better between relatively similar sites and value function transfers produced lower errors between dissimilar sites. For health, Ready and colleagues found that adjusting for differences in health status, income and demographics, by ad hoc scaling or by function transfer, did not improve transfers of willingness to pay to avoid ill-health episodes between European countries.

Robinson and colleagues describe how meta-analysis of valuation studies yields reduced-form equations that transfer values from a few inputs, and advise starting with a systematic review so that only adequate studies of similar outcomes are pooled.

Adjusting for income with an elasticity

Income is the usual adjustment when a VSL is moved between countries, because willingness to pay rises with ability to pay. Robinson and colleagues give the standard formula, shown below.

$$VSL_{t} = VSL_{b} \times \left( \frac{Y_{t}}{Y_{b}} \right)^{\eta}$$

where $VSL_{t}$ is the value at the target (policy) site, $VSL_{b}$ is the base value, $Y_{t}$ and $Y_{b}$ are average incomes at the two sites, and $\eta$ is the income elasticity, the percentage change in value for a 1% change in income. An elasticity of 1 keeps the ratio of value to income constant; above 1, the value falls faster than income in poorer populations.

The elasticity is the decisive choice. Hammitt and Robinson reported that US regulatory agencies typically assume 0.4 to 0.6 for income growth over time, that transfers between high- and lower-income countries often use 1.0, and that several lines of research support values above 1.0. Viscusi and Masterman estimate 0.5 to 0.7 within the United States and just above 1.0 elsewhere. For global benefit-cost analysis, Robinson and colleagues recommend a standardised sensitivity analysis whose preferred default extrapolates from a US VSL-to-income ratio of 160 with an elasticity of 1.5, with a floor of 20 times income per head, alongside constant ratios of 160 and 100 times income (an elasticity of 1.0).

Price level and currency are separate steps. Robinson and colleagues advise inflating a value in its original currency before converting it, and use purchasing power parity rather than market exchange rates because it better reflects the resources each population has. The choice matters most for poorer countries. On the World Bank figures they cite, India's gross national income per head in 2015 was 6,060 international dollars at purchasing power parity but USD 1,600 at exchange rates.

When a transfer is defensible

A transferred value is only as good as the match between the two sites. Circular A-4, issued in 2003 and reinstated in 2025, sets selection criteria that apply directly to health values. Source studies should use adequate data and sound methods and report the parameters of the valuation function. The populations and affected market size should be similar, as should the good, the size of the change, its reversibility, the availability of substitutes and the distribution of property rights, so that the same welfare measure applies.

The property-rights criterion matters in health because willingness to accept and willingness to pay can differ substantially; Circular A-4 states that transfer is not appropriate when the study context supports one measure and the policy context the other. It also rules out transfer for goods with unique attributes, and the use of values from studies of small marginal changes for policies involving large changes. Because transfer is quick and cheap but carries uncertainties and biases of unknown size, the Circular calls it a last resort needing explicit justification. In health, a value for avoiding a mild episode of illness says little about a severe or chronic condition.

Measuring transfer error

A transfer method is tested by comparing a transferred value with a primary estimate made at the policy site itself, a form of convergent validity testing. The usual summary is the percentage transfer error, shown below.

$$TE = \frac{\left| \hat{V}_{p} - V_{p} \right|}{V_{p}} \times 100$$

where $TE$ is the transfer error in per cent, $\hat{V}_{p}$ is the value transferred to the policy site and $V_{p}$ is the value estimated directly at the policy site.

Ready and colleagues ran parallel contingent valuation surveys in five European countries on willingness to pay to avoid ill-health episodes linked to air and water quality; transferring unit values between countries gave an average transfer error of 38%. A source study's confidence interval does not capture this error. Analysis of parameter uncertainty should therefore vary the source value and the elasticity, and analysis of structural uncertainty should test alternative source studies and transfer methods.

How the Green Book handles transferred values

The 2026 edition of HM Treasury's Green Book does not use the term benefit transfer; its closest treatment is of generic prices, set through broadly agreed research and analysis rather than observed in a market. It asks practitioners to review the studies behind generic prices, which may have been calculated in contexts significantly different from the proposal. Where no reliable market or generic price exists, bespoke research proportionate to the proposal may be needed, and because non-market prices are often uncertain, a range of estimates and sensitivity analysis may be required.

For health, the Green Book names the value of a prevented fatality, published in the Department for Transport's TAG data book; the value of a statistical life year; and the value of a quality-adjusted life year in its supplementary guidance on health. These are based on average values from representative population samples and are not intended for emergency or rescue contexts. Applying them to a subgroup or setting unlike that average is itself a transfer.

Worked example with one programme and three transferred VSLs

All figures are illustrative except the reference-case parameters from Robinson and colleagues, which are a US VSL-to-income ratio of 160, US gross national income per head of 57,900 international dollars, and the floor of 20 times income. A screening programme in a middle-income country cuts annual mortality risk by 1 in 10,000 for 500,000 adults, averting 50 expected deaths a year, at an annual cost of 45 million international dollars. Income per head is 10,000 international dollars at purchasing power parity.

1. Elasticity of 1.5 (preferred default). Written as a ratio to income, the value is as follows.

$$\frac{VSL_{t}}{Y_{t}} = 160 \times \left( \frac{10000}{57900} \right)^{0.5} = 66.49$$

where $VSL_{t}/Y_{t}$ is the target-country VSL as a multiple of income per head and the exponent 0.5 is the elasticity minus one. The VSL is 66.49 × 10,000 = 664,900 international dollars, above the floor of 20 × 10,000 = 200,000.

2. Constant ratio of 160 (elasticity 1.0). The VSL is 160 × 10,000 = 1,600,000 international dollars.

3. Constant ratio of 100 (elasticity 1.0). The VSL is 100 × 10,000 = 1,000,000 international dollars.

4. Annual benefits and the decision. Each VSL is multiplied by 50 averted deaths and set against the 45 million cost, giving a net benefit and a benefit-cost ratio.

The programme passes a cost-benefit test under both constant-ratio assumptions and fails under the preferred default. Its health effect is identical in every row; only the base ratio and the assumed income elasticity change. A report using one transferred value would hide that the decision rests on an unsettled base value and elasticity, which is why the reference case asks for all three results.

Limits and common misreadings

Benefit transfer makes a monetised appraisal possible where no primary study exists, but it is easy to give a transferred number more authority than it has. The common errors are listed below.

• Treating the transferred value as precise. A 38% average transfer error is invisible in a source study's reported sampling error.
• Adjusting for income while ignoring the good. Income scaling cannot repair a mismatch in the risk change, its severity or its baseline.
• Mixing units of value. A value per household, per episode or per year applied per person or as a one-off value misstates the total, as does summing component values that a programme value already includes.
• Using national income as if it were respondents' income. Among studies reporting individual income, Robinson and colleagues found sample means from about 29% to about 290% of national income per head, so national averages may introduce substantial error.
• Reading income-adjusted values as statements of worth. A lower VSL in a poorer country reflects lower ability to pay; whether that should lower the priority given to its population's health is a separate, normative judgement for the decision maker.

A transfer is most credible when the study and policy sites value the same change in similar populations, the adjustments are few and reported, and the net present value or benefit-cost ratio is shown across the plausible range of source values and elasticities.

Sources

• Bateman IJ, Brouwer R, Ferrini S, Schaafsma M, Barton DN, Dubgaard A, et al. Making benefit transfers work: deriving and testing principles for value transfers for similar and dissimilar sites using a case study of the non-market benefits of water quality improvements across Europe. Environmental and Resource Economics. 2011;50(3):365-387. https://doi.org/10.1007/s10640-011-9476-8
• Hammitt JK, Robinson LA. The income elasticity of the value per statistical life: transferring estimates between high and low income populations. Journal of Benefit-Cost Analysis. 2011;2(1):1-29. https://doi.org/10.2202/2152-2812.1009
• HM Treasury. The Green Book (2026). London: HM Treasury; 2026. Chapter 8. https://www.gov.uk/government/publications/the-green-book-appraisal-and-evaluation-in-central-government/the-green-book-2026
• Johnston RJ, Boyle KJ, Loureiro ML, Navrud S, Rolfe J. Guidance to enhance the validity and credibility of environmental benefit transfers. Environmental and Resource Economics. 2021;79:575-624. https://doi.org/10.1007/s10640-021-00574-w
• Office of Management and Budget. Circular A-4: Regulatory Analysis. Washington, DC: Executive Office of the President; 2003 (reinstated 2025). https://www.whitehouse.gov/wp-content/uploads/2025/08/CircularA-4.pdf
• Ready R, Navrud S, Day B, Dubourg R, Machado F, Mourato S, et al. Benefit transfer in Europe: how reliable are transfers between countries? Environmental and Resource Economics. 2004;29(1):67-82. https://doi.org/10.1023/B:EARE.0000035441.37039.8a
• Robinson LA, Hammitt JK, O'Keeffe L. Valuing mortality risk reductions in global benefit-cost analysis. Journal of Benefit-Cost Analysis. 2019;10(S1):15-50. https://doi.org/10.1017/bca.2018.26
• Viscusi WK, Masterman CJ. Income elasticities and global values of a statistical life. Journal of Benefit-Cost Analysis. 2017;8(2):226-250. https://doi.org/10.1017/bca.2017.12