The price attribute An Achilles' heel in stated preference exercises by liuqingyan


									Measuring preferences for health care
provision I: The price attribute
An Achilles’ heel in stated preference

Dorte Gyrd-Hansen
University of Southern Denmark & Danish Health
Services Research Institute
Why measure patient preferences?

   Primarily to aid policy making
   Primarily to assist prioritisation
    where the market is not operating
   To assist in establishing which
    programmes should be included in
    the insurance package
This talk will focus on patient
preferences as measured by WTP

   Such preferences potentially feed
    into cost benefit analysis
   Theoretical basis: the potential
    pareto criterion
   EXTREMELY important to
       When comparing costs and benefit in a
        CBA – we must not compare apples
        and pears!!!
Apples and pears….
   When eliciting WTP we must always
    ensure that the marginal utility of income
    of beneficiaries is equal to the marginal
    income of (tax)payers
   (if this does not hold the monetary values
    we are comparing do not translate into
    utilities in the same manner)
   Hence we cannot elicit WTP amongst
    those who are dying, those who are very
    ill or other unrepresentative groups (such
    as those with low income)
Apples and pears…
   This means that we must elicit WTP
        For individuals whose health is not so poor that
         it affects the utility of income
        And for smaller health increments
        In an ex ante (insurance) setting
   And WTP questions must be targetted at
    representative income groups
    (alternatively we must adjust WTP value
Preference elicitation
   Stated preference methods are used to estimate
    the value that people place on health care
   Aim: to guide policy
   Problem: little is known about how people make
    decisions in preference elicitation tasks
   There is substantial evidence that people employ
    heuristics (cognitive short cuts) in order to
    simplify the tasks they are presented with
    (Payne et al, 1993; Kahneman and Tversky,2000;
    Gilovich et al,2002)
Preference elicitation
   Contingent valuation (CV):
       Participants are presented with a scenario which
        represents an improvement over the current state
       Participants are then asked to indicate what their
        maximum WTP is for improvement (open-ended or
        closed ended)
   Discrete choice experiments (DCE)
       Participants are presented with hypothetical choices
        where attributes are systematically varied
       DCEs are designed to determine preferences for
        different aspects of treatment
       It is possible to infer WTP from a DCE study if cost
        is included as an attribute
   Complete and stable?
       Well formed preferences - which are
        unaffected by the task – are assumed
   Where preferences are not complete – it
    is assumed that these can be formed on
    the basis of information and attitudes
    during the task
   Evidence indicates that this process is
    largely driven by heuristics
   Time pressure, task complexity and poor
    motivation may lead respondents to
    adopt very simple preference rules
Fast and frugal heuristics
   Gigerenzer et al (1999) suggest that
    compensatory decision making (i.e. trading off) is
    too complex to underlie normal behavior
   Fast and frugal heuristics involves using simple
    search and stopping rules using the minumum
    amount of information necessary to make the
       Lexicographic preferences: focus on one attribute at
        a time.
       IMPLICATION: Questions the interpretation of
        marginal rates of substitution
       THUS: Important to explore the prevalence of
        dominant preferences
   We may also observe insensitivity to scope…
STORY 1: The insensitivity of WTP to
the size of the good

   Olsen, Donaldson & Pereira (2004) focus
    on two types of scope insensitivity:
    number of patients treated and different
    reductions in the risk of heart attack
   Scope insensitivity is at odds with neo-
    classical theory of consumer behavior
   Non-satiation axiom
   Albeit WTP is not expected to increase in
    proportion with the size of the good, but
    at a diminishing rate (due to income
The study by Olsen, Donaldson &
Pereira (2004)….
   Comparisons of WTP was undertaken both
    across and within samples
   Context: heart operations, cancer
    radiotherapy, helicopter ambulance..
   The results showed no statistically
    significant differences in WTP (tax
    payment) with different sized health
   Casts doubt on the reliability of the WTP
    method as an instrument for comparing
    the value of competing health care
Within respondents design

   The majority of respondents stuck
    to the same WTP value for the
    programme that has twice the
   Among those who did revise their
    WTP, there are two major types of
       100% increase in WTP
       50% increase in WTP (mental budgetting?!)
Scope insensitivity - reasons
   Cognitive limitations – individuals focus on the
    existence of an attribute, not on the number
    which indicates the size of the outcome
   Embedding: the respondent may be expressing
    the valuation of a larger programme, and not the
    specific programme in question
   Moral satisfaction/warm glow: WTP response is
    essentially expressive in nature and only
    indirectly linked to the particular good. A positive
    WTP is a signal, rather than a precise expression
    of valuation
   Thalers (1985) concept of mental accounting – a
    particular kind of budget constraint in which
    consumers divide their income into sub-budgets.
STORY 2: Ordering effect and price

   Kjær T, Bech M, Gyrd-Hansen D, Hart-
    Hansen K, 2006:
   Focus on analysing the impact of attribute
    ordering on the relative importance of the
    price attribute.
   A discrete choice experiment was
    performed in order to elicit psoriasis
    patients preferences for treatment
   Price sensitivity was compared when price
    was presented as first and last attribute,
Ordering effect in DCE

   Farrar and Ryan (1999) found that
    ordering had no significant effect on
    the estimated relative utility
   Scott and Vick (1999) observed an
    ordering effect
   None of these studies included price
    as an attribute
Testing for ordering effect
   No ex ante hypotheses regarding whether
    respondents are most price sensitive
    when the price attribute is placed first or
    last in the choice set
   The order of presentation may affect the
    degree to which the price attribute is a
    dominant variable, we would expect the
    number of respondents exhibiting
    dominant preferences for price to be a
    function of where this attribute is placed
A possible choice scenario:
price attribute presented first
A possible choice scenario:
price attribute presented last
Challenges when testing for effect of

   Respondents were randomised to
    questionnaire variant
   Presenting price first or last may affect
    cognitive burden, and hence scale
    (degree of random error)
   The magnitude of parameter coefficients
    are dependent on scale.
   Hence, when pooling data from the arms
    of the study we must test for differences
    in scale – and correct for this in order to
    make the comparison
Lexicographic preferences

   Lexicographic preferences (take the
    cheapest) more prevalent when
    price is placed last?
   Focus on dominant preferences –
    i.e. respondents consistently
    choosing the cheapest across all 8
    scenarios that they are presented
Analysis of dominant preferences
STORY 3: Do people reactions to
payment incur kinked utility curves?
   Conjoint Analysis. The cost variable: an
    Achilles heel? Skjoldborg US & Gyrd-
    Hansen D.
   Setting: DCE exercise focusing on (choice
    of hospital and) choice of health care
   Model comparisons are performed which
    test the effect of
       The cost range applied
       The effect of including a dummy variable to
        represent the utility asscociated with payment
        per se
The design
DCE and the price attribute
   A wider cost of including higer payments
    is associated with lower parameter value
    of price attribute
   Including a dummy variable to explain
    utility associated with payment par se has
    significant effects on the model
   Results suggest that inclusion of a two-
    dimensional structure to explain the
    relationship between cost and utility may
    avoid erroneous conclusions and give rise
    to significant changes in implicit WTP
          Highest price level omitted for Includes dummy
Results   extra tax payment attribute     Variables =1 if payment
                                          else zero
The importance of the highest bid
   Ratcliffe (2000): ”if the highest level of
    cost is set too low…..individuals would be
    willing to pay more to receive their
    chosen intervention…and the WTP value
    inferred will be underestimated”.
   Our result: 240 choice scenarios involved
    a tax payment of 25000 DKK. In 25 out of
    the 240 cases, the option which involved
    25000 DKK was chosen. The fact that
    10% chose to pay this amount affects
    WTP considerably.
Measuring preferences for health care
provision II: the difficulties in finding unique
(monetary) valuations of quality of life and
length of life improvements
     STORY 1: WTP for a QALY

   Focus is here on the THEORETICAL &
    EMPIRICAL foundation for eliciting a WTP per
    QALY threshold on the basis of the PUBLIC’S
    PREFERENCES for health relative to other
   Such a threshold would determine the welfare
    optimising SIZE OF THE HEALTH CARE
    BUDGET – which would be a function of the
    cost effectiveness of health care interventions
    AND public preferences for health.
   Applying elicited WTP per QALY thresholds
    could be a move towards Platos philosophy
    rule and an elimination of politicians input.
   Is this realistic? Probably not
Introductory comments II
   Not one but several context specific
    societal WTP values would be
    required in order to reflect societal
    preferences for equity
   Elicitation of a societal WTP would
    require that the following type of
    WTP questions are posed:
   “imagine that you are a politician….”
    or “would you vote for proposal X if
    it meant that you were to pay Y
    euros more in tax?”
Introductory comments III

   Underlying the societal WTP per
    QALY is the individual WTP per
   This is the focus of the present talk
   What are the basic problems
    involved in eliciting individuals’ WTP
    for own health improvements?
Background: Why?
   There is a push for eliciting one
    unique WTP per QALY value
   Translation of cost effectiveness
    analyses (CEAs)  cost benefit
    analyses (CBAs)
   It will give us the answer to the
    question of whether an
    interventions is welfare improving
    or not!
Background: Why not?

   CEAs are based on extra welfarism
   CBAs are based on welfarism
   Welfarism ≠ extra welfarism
   Is it at all possible to align the two
       Theoretically?
       Empirically?
The theory

   Welfarism: individual preferences
    for health outcomes relative to
    other arguments in the utility
   Welfarism: Health is a relative
    concept related to the expectations
    and abilities of people. Preferences
    may differ across individuals
    The theory
   Extra welfarism: output of health care is
    NOT defined in terms of preferences for
    health vis-á-vis other goods, but according
    to its contribution to health itself
   Extra-welfarism: seeks to exclude any
    variation across income/social groups in
    utility derived from improvements in
   Reed Johnson (2005): ” ..QALYs and WTP make
    odd bedfellows. WTP per QALY combines a construct
    whose chief virtues are simplicy and convenience with
    a construct that is not particularly simple or
    convenient but is more consistent with its conceptual
The theory
   If we assume that the QALY is a
    valid cardinal utility measure
   …the only difference between
    WTP and QALY is the scale
   …a constant WTP per QALY would
    require that the two numéraires
    (the scale units) are linearly
    Time and money
   If QALY tariffs are based on the time-
    trade-off technique, then a linear
    relationship would require linearity
    between the value of time and the
    value of money
   The less valuable the numéraire is to a
    person – the higher the number
    requires to express his/her net benefit.
   Different marginal rates af substitution
    between life-expectancy and
    income/consumption → WTP per QALY
    will vary even if the utility associated
    with health gains is the same.
Factors that may affect the marginal
rate of substitution

   Health status influences the
    marginal utility of wealth and the
    marginal utility of life-years
   Wealth influences the marginal
    utility of wealth and possibly the
    marginal utility of life years (if life-
    expectancy is a function of wealth)
   The young and poor versus the old
    and rich…
The theory
   Additional problems arise if the
    QALY is not a measure of cardinal
   WTP will vary with personal
    characteristics; if one does not
    allow QALY values to vary - this
    causes variation in WTP per QALY
Factors that may influence the
valuation of health improvements

   Wealth
   Jobtype, level of education
   Age
   Distribution of health gains:
       given decreasing marginal utility of
        health WTP per QALY would vary
        according to the number of QALYs
        gained at the individual level
Maybe we should just leave theory
alone – and be pragmatic…

   Reed Johnson (2005): ”Whatever
    conceptual flaws may be involved,
    the outcome is unlikely to be worse
    than using an arbitrary threshold of
    $50,000 per QALY”.
   Let us look at what the empirical
    literature tells us…
Is pragmatism viable?
Empirical results
   Franic, Pathak, Gafni (2005):
       Post chemotherapy nausea and vomiting
        (acute condition) ; breast cancer treatment
        (chronic condition)
       SG, VAS, WTP
   Results: QALYs was a poor predictor of
    WTP in both settings
   Possible reasons:
    A problem with SG and non-chronic
    A problem with WTP and serious (life-
    threatening) chronic conditions?
    Empirical results
   Little correlation between SG values and
    WTP (O’Brien and Viramontes, 1994 & Bala
    et al, 1998 )
   WTP and TTO values correlate (Smith,
       A within-subject study
   SG scores negatively associated with
    income (Sackett & Torrance, 1978; O’Brien
    and Viramontes, 1994)
   SG scores negatively associated with low
    educational attainment (Tam, 1982)
   Social class was significant in explaining
    between-subject variation in TTO scores
    (Dolan et al, 1996)
Empirical results
   WTP for health improvements. Direct
    elicitation, mean values:
   Gyrd-Hansen (2003): $15,000
   King et al (2005): $12,500-$32,200
    (Virginia; patient with haboring
    aneurisme/spine condition – one off
   Byrne et al (2005): $1,221-$5,690
    (Texas; osteoarthritis – WTP for knee
Empirical results
   WTP per QALY extrapolated from the
    value of life literature (Hirth et al: 2000):
       $184,200 (contingent valuation; 8 studies;
        median value)
   Discrepancy suggests that the presence of
    mortality risk reduction substantially
    increases stated WTP per QALY
   Appropriate threshold for cost
    effectiveness may be dependent on the
    context of the situation, including risk of
The issue of public-private
   VOSL literature most often based on risk
    reductions that are a private responsibility
    (for example air bags in cars)
   WTP for health improvements – perceived
    as a public health care responsibility (in
    some countries)? Free rider problem?
   Is personal income the appropriate budget
    contraint? (Smith & Richardson, 2005) –
    Necessary to apply a societal perspective?
 Insensitivity to scope/ceiling effect
• Some empirical results show no
  relationship between WTP and QALY
  gains. Some show some sensitivity
  to scope
• Large health increments – lower
  WTP per QALY
• Ability to pay a problem when
  valuating larger health
• Ex ante perspective necessary??
    WTP per QALY vary with socio-
   WTP per QALY has been shown to vary
    with age, educational level, income
    level (Gyrd-Hansen (2006);
    Kontodimopoulos & Niklas (2006), King
    et al (2005))
   If thresholds are applied to existing
    CEAs where the socio-demographic
    characteristics are not readily available
    – we cannot take these factors into
   Should we adher to the notion of
    welfarism; or ignore these variations?
   There is no theoretical basis for one unique WTP
    per QALY threshold (due to variations across
    scales, variations in preferences for health)
   Empirical evidence show great variations across
    context and indviduals
   We may choose to ignore these variations…
   But this would not solve all our problems: When
    eliciting WTP per QALY need to think carefully
    about study design and interpretation of results!
    (societal & insurance based settings?)
   We need to be highly critical when interpreting
    the many WTP per QALY studies which are likely
    to surface in the years to come.
STORY 1 continued :
Results of a Danish study

o Aim of study:
  o To test for correlation between
    expressed WTP and QALY gains derived
    from the TTO exercise
   To analyse the impact of socio-
    economic variables on willingness to
    trade money and time
   To estimate WTP per QALY based on
    individuals’ preferences for own health

   A random sample of Danish (+18)
    subjected to personal interviews
   Subjects randomised to one of 6
   Arms differed w.r.t. initial health
    state and magnitude of health

   Individuals were presented with
    TTO exercise(s)
   A standard interview protocol was
   10 year life-span was applied
   Subsequent to TTO exercise
    respondents were asked to state
    whether they felt sure/unsure about
    their response
   Response time was registered
The principle of the TTO exercise

        Time Trade-off
    WTP exercise
   Respondents were subsequently presented
    with ex post WTP questions
   Why ”ex post” and not ”ex ante”
    (insurance based questions)?
      TTO questions are posed ex post!
      Mixing the ex post and the ex ante
       perspective will entail non-linearity.
      Option value is likely to depend on initial
       health state, health improvement and risk
       of becoming ill. In order to impose such a
       value where relevant it should be
       disentangled from the “base” WTP for a
   Problems with ex ante approach:
      Decreasing marginal utility of income with
       lower initial health state.
    WTP exercise
   Respondents were asked to imagine that they were to
    pay out-of-pocket for the health improvement
   Payment on a monthly basis over a 10 year period
   Prior to the WTP questions a cheap talk script was
    presented to respondents
   Respondents were initially presented with a closed
    ended WTP question where respondents were allocated
    to bids of 100 DKK, 250 DKK, 500 DKK, 1000 DKK,
    2500 DKK, 5000 or 10000 DKK
   Subsequently presented with a payment card (22
    different payments ranging from zero to DKK 50,000)
   Open ended WTP responses were applied

   1724 succesful interviews
   Participation rate: 38.6%
   Some responses were excluded
    when objectively inferior health
    states were valuated higher than
    superior health states (n= 110)
   107 individuals declined to answer
    the WTP question
Zero bids across scales

             QALY gain
             >0     =0
WTP     >0   1069   376   1445
        =0     36    26    62
Total        1105   402   1507
 Mean WTP for different levels of QALY

QALY gain        N      Mean (median)
ΔQALY = 0        402    2467 (1000)
0<ΔQALY≤0.25     423    2527 (1000)
0.25<ΔQALY≤0.5   248    3194 (1250)
0.5<ΔQALY≤0.75   169    4061 (1000)
0.75<ΔQALY≤1     85     2288 (1000)
1<ΔQALY          180    3350 (1000)
    WTP per QALY depends on which sub-
    samples we focus on:

   Full sample:                   45,900 DKK
   Excluding QALY gains >0.75:   171,050 DKK
   Only 0 ≤ QALY gains ≤0.25:    405,000 DKK
   Only 0 < QALY gains ≤0.25 :   237,000 DKK
    Estimated QALY gain, log WTP, log WTP/QALY as
    a function of socio-economic factors and health state
                      QALYgain           Log WTP             Log WTP per QALY

Variabel              Coeff.   p-value   Coeff.    p-value   Coeff     p-value

Constant              0.251    0.000     6.058     0.000     10.080    0.000

Young<30              0.053    0.020     0.418     0.003     0.030     0.842

Old >60               0.021    0.295     -0.500    0.000     -0.531    0.000

Schooling <10 years   -0.012   0.549     -0.349    0.007     -0.428    0.002

Further education     -0.009   0.665     0.073     0.559     0.286     0.044

Household income      0.034    0.091     -0.329    0.007     -0.315    0.023
<200000 DKK p.a.
Household income      -0.001   0.963     0.612     0.000     0.478     0.004
>600000 DKK p.a.
    Estimated QALY gain, log WTP, log
    WTP/QALY as a function of socio-
    economic factors and health state (II)
             QALYgain           Log WTP             Log WTP per QALY

Variabel     Coeff.   p-value   Coeff.    p-value   Coeff    p-value
Arm 1        -        0.001     -0.190    0.245     0.199    0.280
Arm 3        0.023    0.392     -0.042    0.803     -0.101   0.580
Arm 4        -        0.000     -0.011    0.947     -0.144   0.455
Arm 5        -        0.001     -0.309    0.085     -0.180   0.391
Arm 6        -        0.001     -0.124    0.466     -0.106   0.583
Unsure TTO   0.045    0.000     n.a.                -0.090   0.214
Unsure WTP   n.a.               0.288     0.000     0.138    0.084
Time TTO     0.000    0.000     n.a.                0.000    0.536
Time WTP     n.a.               0.003     0.000     0.003    0.000
Log of WTP as a function of QALYs gained and
socio-economic characteristics

Variabel              Coefficient   p-value
Constant              5.847         0.000
Young <30             0.401         0.004
Old >60               -0.500        0.000
Schooling <10 years   -0.342        0.005

Further education     -0.074        0.550

Household income      -0.330        0.007
<200,000 DKK
Household income      0.609         0.000
>600,000 DKK
QALYGAIN              0.860         0.001
QALYGAIN +0.75        -3.342        0.001
QALYGAIN +1           2.473         0.003
Non-trader/willing trader as a function of
health state and socio-economic factors

o   Non traders: individuals who are not
    willing to trade life-years
o   Willing traders: individuals who are willing
    to trade more than 7,5 years (out of the
    initial 10)
o   Willing traders significantly more
    prevalent for initial health states that are
    objectively inferior and/or objectively
    large improvements in health
o   Visa versa for non-traders
 Non-trader/willing trader as a function of
 health state and socio-economic factors
                    Non traders             Willing traders
Variabel            Coefficient   P-value   Coefficient   P-value

Constant            -1.203        0.000**   -1.623        0.000**
Young <30           0.943         0.000**   0.616         0.000**
Old >60             0.002         0.992     0.340         0.037*
Schooling <10       0.029         0.872     0.415         0.008**
Further education   -0.044        0.819     0.152         0.360
Income <200,000     -0.182        0.354     0.094         0.538
Income <600,000     0.239         0.246     -0.302        0.142
Arm 1               0.434         0.047*    -0.363        0.128
Arm 3               -0.530        0.041*    -0.095        0.686
Arm 4               -0.657        0.016     0.937         0.000**
Arm 5               -0.479        0.078     0.463         0.044*
Arm 6               -0.422        0.099     0.668         0.002**
   There is a (near) linear relationship between WTP
    and QALYs for QALY gain <0.75
   WTP does not increase proportionately with QALY
    gain due to a relatively high WTP for those who
    express QALY gain = 0
   How do we deal with QALY gain = 0? These
    individuals value health improvements on the
    WTP scale
   There appears to be rationality behind the act of
    non-trading (most prevalent for scenarios where
    initial health is relatively good)
   Variations in WTP per QALY are partly explained
    by socio-economic characteristics: age (-),
    education (+), income (+)
   and time taken to respond to the questionnaire
Story 2: Small for all or gambling for the
prize? (Gyrd-Hansen D, Kristiansen IS)

   In CEA average postponements of
    (death) are used directly as
    measures of health gains
   Assumption: the discounted value
    of (quality-adjusted) life-years
    represent valid measures of benefit,
    and a good basis for social decision

    The use of average health gains as
     a basis for priority setting is based
     on a number of assumptions:
    1.   That individuals are risk neutral
    2.   Individual utility is a smooth and
         concave function of expected value
         (where concavity is a reflection of
         time preferences)
    3.   Individual utility is equivalent to
         social value
The aim of the empirical study

   To test:
       whether Individual utility is a smooth
        and concave function of expected
       whether individual utility is equivalent
        to social value
   A random sample of Danes were
    presented with a hypothetical intervention
    targeted at patients with risk of heart
   Interviews were performed by
    professional interviewers
   Format: a discrete choice experiment
    where respondents were to choose to
    participate or not participate in the
    offered programme.
   Attributes:
       Probability of gain:
        ”one in X patients will gain from the intervention”
        Values of X: 1,6,12,36,72,120,180
       Size of gain
        ”individual who gain will live Y months longer”
        Values of Y: 1,6,12,36,72,120,180
       Cost of programme: constant across alle
   Full factorial design – all combinations presented
   Scenario was presented as social or individual
   Individual framing:
    ” Imagine that your doctor tells that you have an
      increased risk of heart attack. Your doctor offers
      you a medication which is to be taken orally once a
      day. The pill has few and mild side-effects. This
      treatment requires that you attend two control
      visits to your doctor each year. Your annual out-of-
      pocket expense will be DKK 500.
      Your doctor tells you that 1 in [X] of those who take
      mediciation will live Y months longer, while the rest
      of the patients will have no gain of taking the
      medication. Would you choose to take the
   Social framing:
    ” Imagine that your local regional health authority is
      contemplating spending resources on a new
      medication which can prevent heart attacks. The
      medication is to be taken orally once a day and the
      pill has few and mild side effects. The treatment
      requires that patients attend two control visits to
      the doctor each year. The total costs of doctor visits
      and medication will amount to DKK 5000 per
      patient per year.
      1 in [X] of those who take mediciation will live Y
      months longer, while the rest of the patients will
      have no gain of taking the medication.
      Do you think your county should spend the
      resources on this medication?”
   Respondents were randomised to
    individual or social framing
   Respondent were randomised to
    probability of effect and effect size
   Each respondent was presented with 3
    scenarios (all either social of individual
   On average each version (98 in total)
    were presented to 90 respondents
Results I
   The sample of respondents was not
    entirely representative of the Danish
    population: 20-29 year olds and 65-69
    year olds were over-represented
   Participation rate: 42.4%
   2923 succesful interviews
   Individual choice: females and
    respondents with higher level of
    education more inclined to participate
   Social choice: females and respondents
    with low income more inclined to support
Results II : Sensitivity to scale…
Results III: U-shaped function
(for small gains: preferences for spreading)
Results IV: testing for thresholds
    Conclusions I
   The general choice pattern demonstrates that
    respondents are sensitive to information on
    probability of gain and magnitude of gain
   The results of the study indicate that the value of
    the hypothetical treatments is not a simple function
    of expected value
   Respondents seem to adopt thresholds when they
    value treatment offers.
   Holding expected value constant, small gains are
    associated with less utility
   Standard methods in cost-effectiveness/cost-utility
    analysis where focus is on the sum of health gains
    and not on the distribution of health gains may
    misinform decision makers.
Conclusions II
   Results confirm the suggestion by Olsen
    (2000) and the results by Rodriguez-
    Miguez and Pinto-Prades (2002) that
    social values are characterised by
    preferences for both spreading and
    concentration of health gains
   Social and individual choices do not differ
   Our results suggest that we should be
    careful when interpreting societal
    preferences for the spreading of health
    gains as merely expressions of
    preferences for equity.
Story 3: Willingness to pay for a
statistical life in the time of a pandemic

   A study lauched when there was
    much awareness of the threat of
    avian flu
   This study took place in Norway
   Authors: Dorte Gyrd-Hansen, Peder
    Andreas Halvorsen, Ivar Sønbø
Great variation in value of statistical life

   From the evidence extracted from
    housing and product markets VOSL lies in
    the range $0.7-$9.9 million (Viscusi and
    Aldy, 2003)
   What is the VOSL based on WTP for
    Tamiflu in the context of a threat of an
    influenza pandemic?
   Tamiflu: an ameliorative therapy in case
    one contracts influenza – symptom relief;
    less risk of fatal events
Factors influencing estimation of the
welfare equivalent
   One may argue for the use of different VOSLs
    since people do not consider all fatal events to be
    the same (e.g. process of dying) (Sunstein 1997;
    Beattie et al 1998)
   Revesz (1999) argues that one should employ a
    seperate VOSL for cancer deaths to reflect the
    fact that cancer is especially dreaded
   One may argue that an influenza pandemic has
    characteristics that induce fear: it is
    unpredictable, uncontrollable, kills people at any
    age and may disrupt essential societal functions,
    and involves many deaths at one time.
   Related to so-called disaster aversion
    (Zeckhauser, 1996)
Estimation of VOSL
   WTP per individual * 1/absolute risk
   WTP =1000 Euro; ARR=1/1000
   VOSL= 1000*1000 = 1,000,000 Euro
   Problem with ”tailored” VOSLs:
       Perceived risk reduction may be different from
        actual risk reduction. If perceived RR > actual
        RR – then VOSL is underestimated
       Fear may not be a linear function of perceived
        base-line risk.
       Fearful individuals are relatively insensitive to
        risk changes
The study
   Questionnaire sent to an internet-based
   In total 1750 individuals were invited
   The agreed number of responses (1165)
    was obtained only 3 days after the e-mail
   The questionnaire presented 7 questions
    about the risk of a ”serious influenza
The questionnaire…(some questions)
   Mean VOSL was NOK 60.1 million (USD
    9.7 million)
       Is clearly in the upper end of previous
   Median VOSL was NOK 9 million (USD
    1.45 million)
   Fearful individuals (individuals who were
    willing to take strong precautions):
       Mean VOSL NOK 40 million (median NOK 2.24
        million) !
   Fearful individuals:
       Perceive base-line risk and ARR as higher
       Express higher WTP
       Do not focus on ARR when expressing WTP. Fear
        enters as a constant factor
       Express lower VOSL
   Trusting individual (those who express trust in
    health authorities)
       Perceive base-line and ARR as lower
       WTP is no different than the remaining sample
       Have a tendency to expressing higher WTP per ARR
       Express higher VOSL
Overall conclusions
   WTP a troublesome instrument –due to
       insensitivity to scope
       ordering effect (in DCE, may be equally relevant in
       potential thresholds for WTP (including protest bids)
   Measuring WTP for health outcomes is difficult,
       it involves aligning two scales
       Preferences may be affected by threshold:
        minimum gains; minimum probability of gain
       WTP for health may be context specific – and may
        for example be influenced by notions such as trust
        and fear

To top