Discounting

Discounting in Economic Evaluation

Discounting gives a cost or health gain that arrives later a smaller weight than the same amount arriving now, by multiplying the value for each year by a discount factor that shrinks as time passes. In a cost-effectiveness model it turns streams of yearly costs and QALYs into single totals, so options whose consequences fall in different years can be compared on one basis. NICE asks for costs and health effects to be discounted at 3.5% a year in its reference case, with a 1.5% rate considered only in tightly defined circumstances. The choice matters most for prevention, vaccination and one-off treatments with lasting benefit, where most of the gain lies years after the spending. Separate pages on the discount rate, present value and differential discounting treat the rate, the arithmetic and the case for unequal rates in more depth.

Why later costs and health effects carry less weight

Attema, Brouwer and Claxton list several reasons for discounting in economic evaluation. The first is opportunity cost, since resources spent on health care could have been invested elsewhere in the economy at a positive rate of return. Their own illustration uses a 2% bond, under which 100 euros next year is worth 98.04 euros now. The other reasons are pure time preference (impatience), catastrophic risk and growth in consumption, because extra consumption adds less utility to people who are expected to be richer in future.

Attema and colleagues place these components in the Ramsey equation, the usual basis for a social discount rate, and HM Treasury builds the 3.5% social time preference rate in the Green Book from the same parts, as the time preference page sets out.

The discount factor and the present-value formula

Social planners most often use constant, or exponential, discounting, in which the same annual rate applies year after year. Attema and colleagues give the discount factor for period $t$ as one divided by one plus the rate, raised to the power $t$, and note that through compounding an outcome counts for less the further away it lies. The present value of a stream is the sum of each year's amount multiplied by its factor.

$$PV = \sum_{t=0}^{T} \frac{X_t}{(1+r)^t}$$

where $PV$ is the present value, $X_t$ is the cost or health effect arising in year $t$, $r$ is the annual discount rate written as a decimal, $t$ is the number of years from the start of the analysis and $T$ is the final year of the time horizon. Amounts in year 0 have a factor of 1 and are not reduced. In the Green Book's own table the factor at 3.5% is 0.9662 in year 1 and 0.8420 in year 5. For an illustrative cost of GBP 1,000 in year 10, the factor at 3.5% is 0.7089, so its present value is about GBP 708.92.

Two conventions shape the result. The Green Book rate is a real rate, so general inflation is removed from values first and the two rates are never added together. Models must also state when within a year each amount falls, since counting a year's QALYs at its start or its end shifts every factor by one period. The discount rate page sets out these timing and price-basis choices. The alternative of hyperbolic discounting, in which the rate of impatience falls with delay, describes individual behaviour better in many experiments but produces choices that reverse over time, and Attema and colleagues note that almost all guidelines keep constant rates, although the Green Book steps its rates down after year 30.

Why health effects are discounted as well as costs

Attema and colleagues describe discounting costs as fairly uncontroversial, but discounting health has drawn more debate, because some argue health cannot be saved and invested like money. The opposing argument, at the level of a health system, is that health care turns resources into health and those resources can be moved between years, so health effects should be discounted as costs are, and at the same rate. The older consistency and postponement arguments for one shared rate, and the modern case for a lower health rate based on the growing value of health, are set out on the differential discounting page. An evaluation that values health in money, as in cost-benefit analysis, can build a growing value into the money value itself instead of changing the rate.

Rates set by NICE and other guidelines

Guidelines differ in the rate, in whether costs and health share it and in what sensitivity analyses they ask for. The NICE manual PMG36 (section 4.5) sets 3.5% a year for both costs and health effects in the reference case. The HM Treasury Green Book 2026 uses 3.5% in real terms for years 1 to 30, falling after that, and 1.5% for health and life effects because the wealth effect is excluded. The Zorginstituut Nederland guideline (2024 version) uses constant rates of 3% for costs and 1.5% for effects. The Second Panel on Cost-Effectiveness in Health and Medicine in the United States recommended 3% for both, with sensitivity analyses over a reasonable range of rates. All four can change, so the rates shown are those in force in October 2026, and the discount rate page gives the full schedules.

The NICE committee may consider the 1.5% rate only when all three criteria in section 4.5.3 are met, with the further conditions set out on the discount rate page. The exception lowers the rate for both costs and health together, so it is a lower common rate and not differential discounting.

Worked example with a one-off treatment and thirty years of benefit

The numbers in this example are illustrative. A one-off treatment adds GBP 300,000 of cost at the start (year 0) compared with usual care and adds 0.8 QALYs in each of the next 30 years, counted at the end of each year. Because all the extra cost falls in year 0, discounting leaves the cost unchanged and shrinks only the QALYs.

The present value of 0.8 QALYs a year for 30 years uses the annuity form of the formula.

$$Q = q \times \frac{1-(1+r)^{-30}}{r}$$

where $Q$ is the present value of the QALY gain, $q$ is the yearly QALY gain of 0.8 and $r$ is the annual discount rate. The fraction is the sum of the 30 yearly discount factors, which is 30 when no discounting is applied.

At 3.5% the QALYs gained in year 30 count for only about 36% of their undiscounted value, and the ICER rises by about 63% compared with no discounting (GBP 20,400 against GBP 12,500). Moving to 1.5%, as the NICE exception allows for a treatment that restores near-full health for a very long period, lowers the ratio to about GBP 15,600. The same programme can therefore look quite different under the reference case and the exception, and NICE asks for a highly plausible case that the benefit lasts. If the extra cost were instead spread over the same years as the benefit, as for a lifelong drug, discounting at one common rate would shrink both by the same factor and leave the ratio unchanged.

How discounting shapes results and decisions

Discounting changes results most when costs and benefits fall at different times. Prevention programmes, screening and vaccination usually spend early and gain late, so discounting weakens their ratios, while treatments whose costs and benefits arrive together are less affected. A longer time horizon adds years that each carry a smaller weight, so the horizon and the rate need to be read together.

Because the rate is partly a normative choice, guidelines ask for results that show its effect. PMG36 asks for the disaggregated table of costs, QALYs and life years by health state to be presented with and without discounting. The Dutch guideline requires a scenario with no discounting, and the Second Panel asks for reference case results reported both discounted and undiscounted, alongside a sensitivity analysis over a range of rates. A result that changes direction across plausible rates is a finding in its own right for an HTA committee.

Neighbouring ideas and where each applies

Several pages on the wiki deal with parts of the same idea, and each answers a narrower question. They are best read as follows.

• The discount rate is the number used, with its sources, timing conventions and the long-term Green Book schedule.
• Present value and net present value are the outputs of discounting, the second netting costs against benefits valued in money.
• Time preference is one reason for discounting, and only part of the rate set by a guideline.
• Differential discounting applies a lower rate to health than to costs, as the Dutch guideline does.
• Hyperbolic discounting describes how individuals discount, which differs from the constant rates used in appraisal.
• Inflation adjustment converts past prices to a common price year, which the Green Book keeps separate from discounting.

Limitations and common misunderstandings

Discounting rests on judgements about growth, risk and fairness, and a single rate hides them. Attema and colleagues point out that the opportunity cost of health spending and the consumption value of health, which ought to set the rates, lack empirical estimates.

• Discounting is distinct from inflation. The Green Book applies a real rate to values that already have general inflation removed, and warns that adding the two rates together gives a mathematically incorrect result.
• Health can be discounted twice. Time trade-off values are elicited from people who themselves discount future years, which biases utilities downward, and models then discount the resulting QALYs again. Attema and colleagues call this double discounting and note that correction procedures exist.
• Constant rates give distant generations almost no weight. Over very long horizons compounding pushes weights close to zero, which raises intergenerational equity questions that a single rate does not settle.
• The NICE 1.5% rate is an exception. It is a non-reference-case analysis available only when the committee judges all the criteria are met.
• Past costs are not compounded forward by the discount rate. The Green Book states that discounting should not be applied retrospectively to costs and benefits that have already occurred.

Sources

• Attema AE, Brouwer WBF, Claxton K. Discounting in economic evaluations. PharmacoEconomics. 2018;36(7):745-758. doi:10.1007/s40273-018-0672-z
• HM Treasury. The Green Book (2026). London: HM Treasury; updated 5 February 2026. https://www.gov.uk/government/publications/the-green-book-appraisal-and-evaluation-in-central-government/the-green-book-2026
• HM Treasury. Green Book supplementary guidance: discounting. London: HM Treasury; February 2026. https://www.gov.uk/government/publications/green-book-supplementary-guidance-discounting
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Chapter 4, economic evaluation, sections 4.2, 4.5 and 4.10. Published 31 January 2022, last updated 13 October 2026. https://www.nice.org.uk/process/pmg36/chapter/economic-evaluation
• Sanders GD, Neumann PJ, Basu A, et al. Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: Second Panel on Cost-Effectiveness in Health and Medicine. JAMA. 2016;316(10):1093-1103. doi:10.1001/jama.2016.12195
• Zorginstituut Nederland. Guideline for economic evaluations in healthcare, 2024 version. Diemen: Zorginstituut Nederland; 2024. https://english.zorginstituutnederland.nl/documents/2024/01/16/guideline-for-economic-evaluations-in-healthcare