Friction Cost Method

Friction Cost Method for Valuing Productivity Losses

The friction cost method puts a money value on the paid work lost when illness, disability or premature death takes a person out of employment, on the premise that the employer eventually fills the gap with someone else. Lost output is counted only for the friction period, the time an organisation needs to restore production, so a long absence or a death is capped at that period instead of being valued over the rest of a working life. The choice between this method and the human capital approach can change a productivity cost estimate many times over, which matters for any cost-of-illness study and for an economic evaluation taken from a societal perspective. Some HTA bodies require the method, while others leave productivity costs out of the reference case altogether.

Why replacement limits the loss to society

Economists in the Erasmus group in the Netherlands put the friction cost method forward from 1992, and Koopmanschap and colleagues presented it in the Journal of Health Economics in 1995. Pike and Grosse summarise the argument that, when unemployment is high, friction costs can be lower than human capital costs if the marginal productivity of labour is diminishing, if short absences are covered by internal labour reserves, if the worker makes up lost production on return, and if non-urgent work can be cancelled.

The central claim concerns long absences and permanent exits. The CADTH guidelines for Canada (4th edition, 2017) state that under the friction cost approach the productivity loss to society is limited to the time before the person is replaced, and that losses from premature death should not extend beyond the friction period. A replacement who was already in work leaves a vacancy of their own, and Pike and Grosse note that the first vacancy can set off a chain of vacancies, each with its own friction period.

Calculating friction costs for an absence or exit

Pike and Grosse list five inputs, namely when a friction period begins, how long it lasts, how much production is lost during it, the costs of searching for and training replacements, and medium-term macroeconomic effects. Few applied studies include all of them. A general expression follows.

$$C_{FC} = \sum_{i} \left[ \varepsilon \, v \, \min(d_i, F) + R_i \right] + M$$

where $C_{FC}$ is the friction cost, $i$ indexes each absence episode or permanent exit from work, $\varepsilon$ is the elasticity of production with respect to working time, $v$ is the value of one day of paid work, $d_i$ is the number of working days lost in episode $i$, $F$ is the friction period in working days, $R_i$ is the recruitment and training cost when a replacement is hired (zero otherwise), and $M$ is any medium-term macroeconomic cost included.

The elasticity reflects the idea that output does not fall one for one with working time, because colleagues absorb some tasks and some work is postponed. Pike and Grosse report that production lost during the friction period is often assumed to be 80% of gross production. Koopmanschap and colleagues estimated short-term friction costs in the Netherlands in 1990 at 1.5% to 2.5% of net national income, depending on how far short absences reduce output, with medium-term macroeconomic effects of absence and disability reducing national income by a further 0.8%.

For absences shorter than the friction period the cap does not bind. CADTH notes that lost production from short absences may be partly restored when the person returns or by the firm's internal labour resources, so the two methods may give similar estimates. Because friction periods last weeks or months, discounting barely affects the friction cost, whereas it changes human capital estimates for deaths a great deal.

Setting the length of the friction period

The friction period is an empirical labour market quantity based on how long vacancies take to fill, so it differs between countries and over time. The 2024 update of the Dutch costing manual, described by Hakkaart-van Roijen and colleagues, calculates it from national vacancy statistics.

$$F_{NL} = \frac{365}{N_f / N_o} + 28$$

where $F_{NL}$ is the friction period in calendar days, $N_f$ is the number of vacancies filled in a year, $N_o$ is the number of open vacancies at a given point in time, and 28 days is the four weeks the manual adds. The ratio $N_f / N_o$ is how many times the stock of vacancies turns over in a year, so 365 divided by it is the average time a vacancy stays open. With illustrative figures of 1,000,000 vacancies filled a year and 250,000 open at one time, the average vacancy lasts 365 / 4 = 91.25 days and the friction period is 91.25 + 28 = 119.25 days.

Averaged over 2018 to 2022, the manual's friction period is 115 calendar days (16 weeks). Pike and Grosse found friction periods between 6 weeks and 6 months in the studies they compared. They also point out that a worker on sick leave still holds the job, so there is no vacancy until the employer decides to recruit, yet studies typically cap sick leave at the same friction period used for permanent departures.

Worked example with absences, exits and a new treatment

All figures in this example are illustrative. A new treatment is compared with usual care in 100 employed patients over one year. A day of paid work is valued at GBP 200, the elasticity is 0.8, the friction period is 80 working days (about 16 weeks), and hiring a replacement after a permanent exit costs GBP 3,000, while long absences are assumed to be covered by temporary staff at no recruitment cost. Usual care has 30 short absences of 10 working days, 10 long absences of 150 working days and 2 permanent exits. The new treatment has 25, 6 and 1.

1. Friction cost per event. A friction day is worth 0.8 × 200 = GBP 160. A short absence costs 10 × 160 = GBP 1,600, a long absence is capped at 80 × 160 = GBP 12,800, and a permanent exit costs 80 × 160 + 3,000 = GBP 15,800.

2. Friction cost per arm. Usual care costs GBP 48,000 for short absences, GBP 128,000 for long absences and GBP 31,600 for exits, a total of GBP 207,600. The new treatment costs GBP 40,000, GBP 76,800 and GBP 15,800, a total of GBP 132,600. The difference is GBP 75,000, or GBP 750 per patient.

3. Human capital cost. Every lost day counts at the full GBP 200, so a short absence costs GBP 2,000 and a long absence GBP 30,000. For a permanent exit, 8 working years remain at 225 days a year, or GBP 45,000 a year, discounted at an illustrative 3.5% with each year counted at its end.

$$PV = 45{,}000 \times \frac{1 - 1.035^{-8}}{0.035} = 45{,}000 \times 6.87396 = 309{,}328$$

where $PV$ is the present value in GBP of the earnings lost through one permanent exit, 0.035 is the discount rate and 8 is the number of remaining working years.

4. Human capital cost per arm. Usual care costs 60,000 + 300,000 + 618,656 = GBP 978,656 and the new treatment 50,000 + 180,000 + 309,328 = GBP 539,328. The difference is GBP 439,328, or GBP 4,393.28 per patient.

5. Where the gap comes from. The usual care human capital total is 978,656 / 207,600 = 4.71 times the friction total. The two exits make up 63.2% of the human capital cost but only 15.2% of the friction cost, so the method matters most for events with the most working time lost.

6. Effect on a societal result. The treatment is assumed to add GBP 6,000 in healthcare costs per patient and to gain 0.20 QALYs. Without productivity costs the ICER is 6,000 / 0.20 = GBP 30,000 per QALY. With the friction cost saving it is 5,250 / 0.20 = GBP 26,250, and with the human capital saving 1,606.72 / 0.20 = GBP 8,033.60. The valuation method moves the result far more than adding friction costs does.

Where HTA bodies stand on the method

The Zorginstituut Nederland guideline for economic evaluations in healthcare (2024) includes productivity costs in the reference case and states that the friction cost method must be used to value productivity losses, with reference prices from the costing manual. CADTH (2017) uses a publicly funded health care payer perspective for the reference case, and for productivity losses from absence from paid work under a societal perspective it names the friction cost approach as the recommended method.

The NICE technology appraisal and highly specialised technologies guidance manual (PMG36) sets the reference case perspective on costs as the NHS and personal social services and states in section 4.2.9 that productivity costs should not be included. Section 4.4.23 allows them to be presented separately where they may be a critical component of a technology's value, so the choice of method arises only in that supplementary analysis. Pike and Grosse report that 51 of the 80 cost-of-illness studies they found using the method came from Canada, Germany and the Netherlands, the three countries that had officially endorsed it as of 2012.

How the choice changes results

All 38 studies in the comparison table of Pike and Grosse gave smaller productivity costs with the friction cost method, but the gap varied widely. The three studies with human capital estimates less than twice the friction estimates all measured short-term sick leave, while all nine with ratios of 19 or more included premature mortality or long-term disability. One rheumatoid arthritis study with a 58-day friction period reported a ratio of 6.3, and another with a 6-month period reported 1.5 to 1.6.

The gap can reverse a conclusion. Pike and Grosse cite a societal analysis of infliximab against prednisone in rheumatoid arthritis that reported EUR 22,000 per QALY with human capital valuation and EUR 131,000 per QALY with friction costs, and suggest that payers may prefer the friction cost method because fewer therapies then appear cost effective. A scenario analysis using the other method, with sensitivity analysis on the friction period and elasticity, shows how far a result rests on this choice.

Criticisms and the reply from its authors

Johannesson and Karlsson argued that the method rests on assumptions not supported by neoclassical theory, and that applying it consistently would also reduce the estimated costs of healthcare programmes. Other critics, reported by Pike and Grosse, doubt that sick workers or internal reserves could make up lost work unless firms were not maximising profit, or hold that the method in effect sets the shadow price of labour to zero.

A further criticism was that the method treats leisure as worthless. Brouwer and Koopmanschap replied with a time-allocation model in which leisure is valued in QALYs, as usual in economic evaluation, and concluded that changes in unpaid work and leisure need valuing separately under either method.

Friction cost, human capital and productivity cost

Productivity cost is the quantity being estimated, and the two methods are ways of valuing it. The human capital approach values potential production lost over the whole absence or remaining working life, while the friction cost method values the production lost before output is restored. Pike and Grosse note that the human capital approach does not assume full employment, since it applies age-specific employment rates, so the methods differ mainly in what happens once a replacement is possible.

The friction cost method concerns paid work, and time away from it is measured as absenteeism. Unpaid work and informal care need their own valuation. Whether productivity costs enter the analysis at all is a question of perspective, settled before any valuation method is chosen.

Common errors in applying the friction cost method

Most problems come from undocumented inputs or from mixing methods within one estimate. Reviews of practice have found both.

1. Borrowing a friction period without a source. Kigozi and colleagues found that only 16 of 46 economic evaluations using the method stated where their friction period came from.
2. Leaving the elasticity unreported. Only nine of those 46 reported the elasticity correction, yet 0.8 against 1.0 changes every friction day by a fifth.
3. Mixing methods across components. Pike and Grosse excluded a study that applied the friction cost method to mortality costs and human capital valuation to unemployment costs.
4. Assuming the result carries across jurisdictions. A friction cost saving counted in the Dutch reference case is excluded from the PMG36 reference case.

Sources

• Brouwer WBF, Koopmanschap MA. The friction-cost method: replacement for nothing and leisure for free? PharmacoEconomics. 2005;23(2):105-111. https://doi.org/10.2165/00019053-200523020-00002
• CADTH. Guidelines for the economic evaluation of health technologies: Canada. 4th ed. Ottawa: CADTH; March 2017. https://www.cda-amc.ca/sites/default/files/pdf/guidelines_for_the_economic_evaluation_of_health_technologies_canada_4th_ed.pdf
• Hakkaart-van Roijen L, Peeters SB, Reckers-Droog V, Evers S, Wijnen B, Huis in 't Veld LM, Thielen FW, Kanters TA. From paper to platform: updating the Dutch costing manual and launching a web application. PharmacoEconomics. 2026;44(6):727-740. https://pmc.ncbi.nlm.nih.gov/articles/PMC13186881/
• Johannesson M, Karlsson G. The friction cost method: a comment. Journal of Health Economics. 1997;16(2):249-255. https://doi.org/10.1016/S0167-6296(97)00006-4
• Kigozi J, Jowett S, Lewis M, Barton P, Coast J. Estimating productivity costs using the friction cost approach in practice: a systematic review. European Journal of Health Economics. 2016;17(1):31-44. https://doi.org/10.1007/s10198-014-0652-y
• Koopmanschap MA, Rutten FFH, van Ineveld BM, van Roijen L. The friction cost method for measuring indirect costs of disease. Journal of Health Economics. 1995;14(2):171-189. https://doi.org/10.1016/0167-6296(94)00044-5
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022, last updated 13 October 2026. Sections 4.2.9 and 4.4.23. https://www.nice.org.uk/process/pmg36
• Pike J, Grosse SD. Friction cost estimates of productivity costs in cost-of-illness studies in comparison with human capital estimates: a review. Applied Health Economics and Health Policy. 2018;16(6):765-778. https://pmc.ncbi.nlm.nih.gov/articles/PMC6467569/
• Zorginstituut Nederland. Guideline for economic evaluations in healthcare. 2024 version. https://english.zorginstituutnederland.nl/site/binaries/site-content/collections/documents/2024/01/16/guideline-for-economic-evaluations-in-healthcare/guideline-for-economic-evaluations-in-healthcare.pdf