Economic evaluation

20th November 2003

Topic 8. Discounting

  1. The rationale for discounting
  • Time preference

 revealed inter-temporal preferences in the marketplace

 the consumption rate of interest to the extent that the public programme displaces consumption

  • Social opportunity cost

 the marginal return on private investment, to the extent that that the public programme displaces private investment.

“Whatever the merits of these arguments, an overarching practical difficulty with any extramarket process for determining the social discount rate is the absence of a well-define d political process or some other mechanism for determining the rate” (Lipscomb et al 1996)

2. Should the discount rate used for benefits be the same as for costs?

  • Objections (and counterarguments) to discounting health outcomes (expressed in quantities) in CEA and CUA.
  • Weinstein and Stason (1977): the consistency argument

Programme / Cost / Benefit
A / $10,000 now / 1 LYG in 40 years’ time
A1 / $70,000 in 40 years’ time / 1 LYG in 40 years’ time
A2 / $70,000 now / 1 LYG now
A3 / $10,000 now / 1/7th of LYG now
B / $10,000 now / 1 LYG now
  • Keeler and Cretin’s paralysing paradox.

Consider 3 options, which are identical except that they are implemented in different time periods. Costs are discounted, but not QALYs gained. Each is appraised in t0

time periods
t0 / t1 / t2…
Option 1 / Costst0/QALYst0
Option 2 / Costst1/QALYst0
Option 3 / Costst2/QALYst0.

For any ‘desirable’ option, the cost effectiveness ratio can be improved by delaying its start date, with the result that attractive options will be indefinitely delayed.

  • A critique of the paralysing paradox

Parsonage and Neuberger (1992)

 The time horizon for budget allocation is finite.

 may be appropriate to use different r for costs and effects.

Lipscomb et al (1996): the Keeler Cretin paradox adds force to the argument for discounting costs and benefits, and at the same rate.

  • Is time preference for health different?
  • Empirical evidence on individual time preference.

“…

  • Individual discount rates frequently lie outside the conventional 0-10% range. In the experiment by Redelmeier and Heller, over 62% of the estimated rates were equal to 0, 10% were less than 0, and nearly 16% were greater than 10%. Some of the discount rates determined by Chapman and Elstein were in excess of 200%.
  • Despite the variation between individuals, the mean rates in many experiments do fall within the conventional range. In the Redelmeier-Heller study, the grand mean was 3.3%. Moreover, the various rates estimated (econometrically) by Viscusi and Moore fell in the range 1-14%.
  • Discount rates tend to be lower when large-magnitude outcomes are veing traded over time (and conversely).
  • Discount rates tend to be lower the longer the rtime interval over which the trades are considered (Cropper et al 1994)…
  • Discount rates for losses are typically lower than for gains.
  • When a given outcome is embedded in a sequence of outcomes, the discount rate tends to be lower than when the outcome is evaluated singly (and thuds not in the context of its outcome sequence).
  • The sequencing of outcomes can affect time preference: some people ‘savor’ good outcomes and wish to postpone them; some ‘dread’ bad outcomes and wish to get them over with; some attach special utility to having outcomes improve over time, even if this means the total payoff (e.g. money) is suboptimal.

In sum, there is much behavioural evidence that individual preference are not consistent with the constant rate exponential discounting model…”

Source: Lipscomb et al (1996) p. 227-8.

3. Functional form

  • Constant rate discounting using a geometric functional form

1(1+ r) t

  • ‘slow’ discounting:

a(t) = [b/(b + t)]

where b > 0.

The relative importance attached to the difference between any two outcomes tends to recede as the outcomes recede into the future.

e.g. individual discount rates are higher when the time interval for making tradeoffs is shorter and the respondent is older.

Problem: as the future time period approaches, the rate of discount between that period and a more distant future time period increases.

 not consistent with the idea of stability of preferences.

4. Pragmatics: how to proceed?

  • US Washington Panel recommendations

 discount costs and outcomes at the same rate.

 use r = 3% but also r = 5% for comparability; sensitivity analysis from 0-7%.

UK:

  • HM Treasury requirements

r = 6%

uncertainty and discounting

new draft of the ‘green book’ proposes r = 3.5%

different rates for £ and Qs

  • NICE guidelines to the preparation of evidence by manufacturers and sponsors

r= 6% for costs, r = 1.5% for outcomes.

Sensitivity: 6% costs, 6% outcomes; 6% costs, 0% outcomes.

New rate for discounting costs = 3.5%; unknown as yet what the corresponding discount rate will be for outcomes.

[urls provided on the topic handout]