VSL
Valuation of Mortality Risks Under The Provision of Contextual And ProductRelated Information Anna Alberini (FEEM and University of Maryland) Milan Ščasný (Charles University Prague)
Workshop on "Recent Trends in Non Market Valuation“ International Center for Climate Governance, Venice, November 3-4
Acknowledgement EXIOPOL - EU FP7 funded project (2007-2011) Task on Mortality valuation Partners: FEEM, Charles Uni Prague, Uni Bath
Motivation • Valuation of benefits of policies that reduce mortality risks • Premature death (lives ‘saved’) relevant for e.g. accidents and jobrelated fatalities Æ VSL multiplied by the number of premature death • “…most policies actually extend lives rather than save them . . . with the resulting change in life expectancy ranging anywhere from 1 year or less to upwards of 30 years” (Nathalie Simon)
• “… even though the impact of air pollution is frequently reported in terms of a number of premature deaths, the appropriate metric is the loss of life expectancy“ (Miller and Hurley, 2003; Desaigues et al., 2011). – The magnitude of the loss of LE per death is very much shorter for premature death due to air pollution [months] than for fatal accidents [30-40y] (Rabl 2003) – By contrast to fatal accidents, the total number of premature death attributable to air pollution is not observable – i.e. the cohort studies cannot distinguish whether a few suffer from large LE loss or everybody loosing a little (e.g. Pope et al., 2002)
Motivation: the impact on mortality Ideally, the effects on mortality risk can be characterised as shifts in individual (population) survival curves. The impact of mortality can be expressed in two ways as a] a change [reduction] in life expectancy [∆LE], or b] a change [increase] in the probability of dying prematurely [∆RISK ] while a] is given as a difference between two integrals of survival function (1 minus b]) thorough one’s life Total benefits in population of N units is typically valued as a] N * ∆LE * VOLY b] N * ∆RISK * VSL
LE gain valuation review Johannesson and Johansson (1997) derive VOLY from WTP for a medical treatment that will increase expecting remaining length of life from 10 to 11 years when she will be 75 Soguel and van Griethuysen (2000) determined a VOLY for air pollution indirectly by asking to rank relative importance of life years lost, mild morbidity (RAD) and severe morbidity; no direct question was however asked about risk reduction or life expectancy Morris & Hammitt (2001) offered to one half of their respondents a life expectancy gain and to second one a reduction of the risk of dying and find that median WTP for the former is higher than median WTP for latter in the group that will take a vaccine in their 60 Desaigues et al. (2004; 2007) on quite small samples experimented with several WTP question variants including a) LE gain of 1,3 and 12 month, b) LE gain corresponding to ∆R=5/1000, and c) showing LE gain as well as ∆R=5/1000.
LE gain valuation: A review /2 VOLY derived from VSL assuming that VSL is the sum of discounted annual VOLY over 30 to 40 years (e.g. ExternE, 2009) Chilton et al. (2004) derive VOLY from the WTP, for the rest of their life, for 1, 3 and 6 months of extra life in normal health as well as in poor health Desaigues et al. (2007; 2011) asked for WTP for 3 and 6 month of life extension when the LE gain is not a matter of additional months of misery at the end of one’s life, but is described rather as pollution causes accelerated aging
Source: Desaigues et al. (2007; 2011)
Motivation: Which impact metric? VSL x VOLY affects the apparent merits of regulatory program that disproportionately affect people with different life expectancies (and age) – benefits of elderly may be much larger using a VSL than a VSLY approach (Hammitt 2007) Æ VOLY called as "senior death discount" (due to shorter expected lifespans of elderlies) – constant VOLY assumes that the VSL is strictly proportional to remaining LE (that is unwarranted) VOLY in use? – US EPA (1999), but only as a second alternative (Desaigues et al., 2004) – ExternE (1996-2009) but VOLY derived from VSL – ExternE (2009 on) based on Desaigues et al. (2007; 2011) – recently, the US EPA‘s SAB rejects using the VOLY approach, similarly Pearce et al. (2006; in OECD CBA) recommends to use VSL rather than VOLY approach
Our research goal • Check consistency across VSL and VOLY by doing SP study with split samples • to elicit the WTP for specified risk reductions, derive the implicit VOLY from VSL, and compare such VOLY with an estimate of the VOLY based on WTP for gains in life expectancy 1. value mortality risk reductions Æ VSL 2. value gains in life expectancy corresponding to the mortality risk reductions in 1. Æ VOLY 3. value mortality risk reductions with reminder of corresponding life expectancy gains Æ VSL* (implied VOLY*)
Revised research goal • Respondents were unable to grasp the notion of gains in life expectancy correctly in the pre-survey – people tend to think of an extension of their life by some months or years, while – the LE is an integral of the survival function thorough one’s life • Analyze the effect of reminding the LE gain corresponding to several profiles of the mortality risk reductions Æ SP study with split samples 1. Value mortality risk reductions Æ VSL 2. Value mortality risk reductions with reminder of corresponding life expectancy gains Æ VSL* • Other split sample treatments and methodological issues to study
Motivation: Contextual VSL (already studies) Subjective perception of the risk – dreaded risk, controllability, exposure, salience etc. – e.g. Alberini and Ščasný (FEEM WP 2010; under review in JRU) Mode of delivery – public program vs. private action – e.g. public premium analysed in Alberini and Ščasný (2010; 11) The cause of death – VSL may vary for specific cause which risk were reduced – e.g. Alberini and Ščasný (2011) use a choice experiments to derive VSL’s for cancer, respiratory illness and road traffic accidents related risks within VERHI-Children project
Our research goal (in the Exiopol project) • whether the cause of death on the VSL matters (all cases, cancer, cardio- and cerebro-vascular and respiratory illness) H0: cancer premium • whether environmental context matters (environmental exposures v. other) H0: VSL larger for enviro context
• whether the duration of risk reduction matters (i.e. the risk reduced permanently (i.e. sustained reductions) or during certain period (i.e. the blip)) H0: WTP for permanent risk reduction is larger than is implied by a sequence of the blips All in private good context (no public program as e.g. in VERHI project)
Choice Attributes and Their Levels Size of risk reduction
2, 3, 4, and 5 in 1000 per decade
Latency
0, 2, 5, 8 years if blip, then the risk reductions last for one decade;
Duration
Cost
if permanent, the risk reduction lasts 4 decades (if the respondent is aged 40-49), or 3 decades (if the respondent is aged 50-60) annual for the next 10 years, starting this year. The amounts: 250€, 500€, 1000€, 1800€, 3000€ (and their eq. in pounds and Czech crowns in PPP).
Choice set
Risk Communication
•
risks expressed as X in 1000 over 10 years (=X in 10,000 per year) A short probability tutorial plus a quiz whether the respondent grasped the concept (select a person with higher prob of dying)
Exper Treatment: LE gain reminder The effect of couching the risk reduction in terms of LE gains v1: risk reduction only v2: with the reminder of expected life time for that person v3: with the reminder on LE gain that is (3x) larger than v2 • …note that we did not use the expression “life expectancy” gain. We just showed current life expectancy (e.g., 79.7 years), and then the LE under each alternative. • Expected life time gain computed from life tables considering person age and risk reduction attributes (i.e. latency and the size of risk reduction). But use same for gender and countries.
Life Expectancy Gain Reminders [used in Treatment 1, version 2] Blip Latency (iin years) 0 2 3 8
dR=2 3 weeks 3 weeks 2 weeks 2 weeks
dR=3 4 weeks 4 weeks 3 weeks 3 weeks
dR=4 6 weeks 5 weeks 5 weeks 4 weeks
dR=5 7 weeks 7 weeks 6 weeks 5 weeks
Permanent Latency (iin years) 0 2 3 8
dR=2 6 weeks 5 weeks 5 weeks 4 weeks
dR=3 2.1 months 8 weeks 7 weeks 6 weeks
dR=4 2.8 months 2.5 months 2.1 months 8 weeks
dR=5 3.5 months 3.1 months 2.6 months 2.4 months
Choice set (with the reminder of expected life time)
Split Sample Treatments Treatment 1: couching of risk reduction [3] (with and without associated life expectancy gains) Treatment 2: cause of death [3] (all, cardiovascular + respiratory, cancer) Treatment 3: environmental exposures [2] (the risks being reduced are associated with environmental exposures vs. no such language) Treatment 4: repeated questioning [2] best/worst v. best/second best
Treatment 5: CAWI vs. CAPI [2] only in the Czech Republic
Survey Administration and Sampling Plan Three countries—Italy, the UK, and the Czech Republic Respondents must be – 40-60 years old – 50:50 men-women – Representative for education and income COUNTRY
MODE and PLACE
SAMPLE SIZE
Italy
CAWI (internet panel, on-line survey), 7 cities
2369
UK
CAWI (internet panel, on-line survey), 7 cities
2426
Czech Republic
CAWI (internet panel, on-line survey), 5 cities CAPI (computer assisted personal interview, at people’s homes), 5 cities + rest of the country
895 2367
The Data ITALY
UK
CZECH
Male, %
49.4
51.1
47.7
Age 40-49, %
50.3
50.0
55.3
Age 50-60, %
49.7
49.6
44.4
32,392
39,277
22,065
20.0 5.8
26.9 9.7
21.2 8.1
29.1
24.5
16.2
Hhold_income, PPP euro Own health, % (1=excellent, 2=very good) (5=poor)
Self-assessed own LE, % (longer than the average person)
Percent responses to the conjoint choice questions by treatment 1 (life expectancy gain)
Response option
1=no LE reminder
2=true LE gain
3= larger LE gain
Alt. A
35.7
24.8
27.7
Alt. B
38.4
30.1
31.9
Neither (status quo)
25.9
45.1
40.3
Random Utility Model D is the marginal utility of a unit risk reduction 'R is the risk reduction per year, G is the discount rate and L latency
Indirect utility depends on discounted flow of risk reductions and residual income; we assume that people do not discount money
D 'R e G L [1 PERM (e G 10 e G 20 AGE 40 e G 30 )] E ( y C) H
V
Eis the marginal utility of income y is income C is the payment
This last term applies only to respondents aged 4049, i.e. AGE40=1
Econometric Model and the VSL
Conditional Logit VSL computed as
VSL for cancer etc.
VSL VSL
(D / E ) 1000 (D1 D 2 ) / E *1000
Results> cause of death ITALY coeff. t. stat ALPHA1 ALPHA2 ALPHA3 BETA DELTA
6.352 0.286 0.342 -17.507 3.327
11794 -12778.0
N log L
all causes cancer cardioresp.
V
0.5864 0.0224 0.0268 -0.0003 0.0288
std. err. VSL (m€) (m€) 2.215 0.304 2.300 0.317 2.317 0.324
UK coeff. t. stat
CZ--CAWI coeff. t. stat
0.263 0.0195 0.1587 -0.0003 -0.0034
4.297 1.1772 0.39 -0.4379 2.556 -0.3529 -21.307 -0.0004 -0.334 0.078
12084 -12976.5
4470 -4735.3
VSL std. err. (m€) (m€) 0.818 0.173 0.878 0.179 1.311 0.260
5.198 -2.13 -1.721 -16.017 3.825
VSL std. err. (m€) (m€) 2.734 0.481 1.717 0.437 1.915 0.418
CZ--CAPI coeff. t. stat 0.8714 0.3071 0.2168 -0.0004 0.079
7.206 2.414 1.728 -26.404 6.924
11835 -12561.0 VSL std. err. (m€) (m€) 2.024 0.257 2.737 0.292 2.527 0.270
(D1 D 2 CANC D 3 CARDIO) 'R eG L [1 PERM (eG 10 eG 20 eG 30 )] E ( y C ) H
Results from main model. All countries.
All countries coeff. ALPHA1 BETA DELTA
no LE gain reminder
ALL
0.6088
t stat. 12.259
coeff. 1.9424
t stat. 19.858
LE reminder==2 coeff. 0.0301
t stat. 1.334
LE reminder==3 coeff. 0.2924
t stat. 4.407
-0.0003 -40.178 -0.0004 -25.594 -0.0004 -27.076 -0.0003 -22.792 0.0344
6.151
0.0709
14.66 -0.0578
-2.164 -0.0015
N
40183
13100
13594
13489
VSL (mEuro)
1.771
5.178
0.085
0.874
s.e. (mEuro)
0.125
0.229
0.062
0.178
-0.132
Results from main model. All countries.
All countries coeff. ALPHA1 BETA DELTA
no LE gain reminder
ALL
0.6088
t stat. 12.259
coeff. 1.9424
t stat. 19.858
LE reminder==2 coeff. 0.0301
t stat. 1.334
LE reminder==3 coeff. 0.2924
t stat. 4.407
-0.0003 -40.178 -0.0004 -25.594 -0.0004 -27.076 -0.0003 -22.792 0.0344
6.151
0.0709
14.66 -0.0578
-2.164 -0.0015
N
40183
13100
13594
13489
VSL (mEuro)
1.771
5.178
0.085
0.874
s.e. (mEuro)
0.125
0.229
0.062
0.178
-0.132
Results from main model mean VSL (std.err.) in million Euro no LE gain reminder
ALL Italy
UK Czech CAWI Czech CAPI ALL
LE reminder
LE reminder X3
2.273
5.869
0.415
1.027
0.264
0.504
0.265
0.391
0.878
6.172
0.195
0.175
0.468
0.125
2.183
4.295
1.096
1.406
0.354
0.366
0.776
0.535
2.425
4.295
1.495
1.924
0.215
0.366
0.519
0.343
1.771
5.178
0.085
0.874
0.125
0.229
0.062
0.178
Responses to the conjoint choice questions by treatment 2 (environmental exposure) Response option
1=mention of 0=no mention environmental exposures
Alt. A
28.6
30.1
Alt. B
32.7
34.2
Neither (status quo)
38.7
35.8
Results: environmental context. All countries
All countries
no envir. context
envir. context
coeff.
t stat.
coeff.
t stat.
0.408
6.35
0.8084
11.139
BETA
-0.0003
-27.827
-0.0004
-28.965
DELTA
0.0143
1.584
0.0485
6.941
N
20170
20013
VSL (mEuro)
1.213
2.306
s.e. (mEuro)
0.169
0.177
ALPHA1
The VSL difference is significant at 1% level
Results: environmental context, mean VSL (s.e.)
no enviro context
ALL Italy Czech CAWI Czech CAPI ALL (CZE,ITA,UK)
with enviro context
2.273
1.918
2.642
0.264
0.38
0.368
2.183
2.201
2.173
0.354
0.453
0.557
2.425
1.687
3.066
0.215
0.332
0.284
1.771
1.213
2.306
0.125
0.169
0.177
Results Responses pass the scope test—except for first version of the life expectancy treatment (v2)
Responses consistent with the economic paradigm – BETA – MU of income positive and significant – ALPHA – MU of the risk reduction is positive and significant VSL for both LE gains are much lower and not different one from the other Low discount rate—but suggests that people do not value future risk reduction as much, even if permanent
But… that model doesn’t cater to the possibility that people might care more for sustained risk reductions than for a sequence of blips. Revised indirect utility functions Æ allows different marginal utility of risk reductions for blip (D1) and sustained risk reductions (D2)
V
'R e
G L
[D1 BLIP D 2 PERM (e
G 10
e
G 20
e
G 30
E ( y C) H … compare to
D 'R e G L [1 PERM (e G 10 e G 20 AGE 40 e G 30 )] E ( y C) H
V
)]
Results: different MU of dR for blip v. permanent Only sample with no mention of LE gains (treatment1, v1) ALL coeff t stat alpha1 alpha2 beta delta
blip permanent
V
ITALY coeff t stat
UK coeff t stat
Czech Rep coeff t stat
1.9752
21.113
1.8917
11.027
2.1889
12.960
1.9439
12.673
1.2113
10.18
1.4836
6.212
1.4789
6.504
0.8914
5.289
-0.0004
-26.395
-0.0003
-12.277
-0.0004
-13.637
-0.005
-19.286
0.0279
4.063
0.0401
3.205
0.0268
0.011
0.0219
1.763
VSL (mill. Euro) 5.029 3.084
VSL (mill. Euro) 5.802 4.550
VSL (mill. Euro) 5.978 4.039
VSL (mill. Euro) 4.128 1.839
'R e G L [D1 BLIP D 2 PERM (e G 10 e G 20 e G 30 )]
E ( y C) H
Results: …with an alternative-specific intercept for sustained risk reductions (D2PERMANENT) Only sample with no mention of LE gains (treatment1, v1) ALL coeff alpha1 alpha2 beta delta
VSL based on alpha1
ITALY
t stat
1.9818 0.2161 -0.0004 0.0962
19.570 5.525 -25.600 12.051
coeff
t stat
1.9436 0.2936 -0.0003 0.0948
10.360 4.024 -12.140 6.243
UK coeff
Czech Republic
t stat
2.2418 0.2671 -0.0004 0.1083
12.005 3.527 -12.255 7.039
coeff
t stat
1.8921 0.1149 -0.0004 0.0851
11.878 1.984 -18.602 7.923
VSL in mill. € (s.e.)
VSL in mill. € (s.e.)
VSL in mill. € (s.e.)
VSL in mill. € (s.e.)
5.297 (0.240)
6.128 (0.546)
6.402 (0.502)
4.262 (0.304)
D1 'R e G L [1 PERM (e G 10 e G 20 AGE 40 e G 30 )] PERM D 2 E ( y C ) H
V
Summary and conclusions SP study that looked at permanent v. blip within the context of reminding people of the LE gains implied by mortality risk reductions Results satisfy scope test and economic paradigm
Reminding people of the LE gains implied by a mortality risk reduction reduces their WTP, and hence the VSL VSL under environmental context not significantly different for country samples People are willing to pay more for a sustained risk reduction than for a sequence of blips, but WTP is only slightly larger
All results so far being presented based on using only the response to the first conjoint choice question for each A-B-status quo choice task. Analysis based on repeated responds (best/best, best/worst) from the same survey will come tomorrow
VSL for CAWI vs. CAPI in Czech Republic
Sub-sample
CAPI
CAWI
ALL
2.48
2.25
Cities (5)
2.55
2.25
Prague
4.02
3.85
Cities (4)
1.79
1.37
age40-49
2.33
2.44
age50-60
2.76
1.94
Contacts: Dr. Milan Ščasný Charles University Prague, Environment Center [email protected] Dr. Anna Alberini Fondazione Eni Enrico Mattei & University of Maryland, USA [email protected]
Workshop on "Recent Trends in Non Market Valuation“ International Center for Climate Governance, Venice, November 3-4
Acknowledgement EXIOPOL - EU FP7 funded project (2007-2011) Task on Mortality valuation Partners: FEEM, Charles Uni Prague, Uni Bath
Motivation • Valuation of benefits of policies that reduce mortality risks • Premature death (lives ‘saved’) relevant for e.g. accidents and jobrelated fatalities Æ VSL multiplied by the number of premature death • “…most policies actually extend lives rather than save them . . . with the resulting change in life expectancy ranging anywhere from 1 year or less to upwards of 30 years” (Nathalie Simon)
• “… even though the impact of air pollution is frequently reported in terms of a number of premature deaths, the appropriate metric is the loss of life expectancy“ (Miller and Hurley, 2003; Desaigues et al., 2011). – The magnitude of the loss of LE per death is very much shorter for premature death due to air pollution [months] than for fatal accidents [30-40y] (Rabl 2003) – By contrast to fatal accidents, the total number of premature death attributable to air pollution is not observable – i.e. the cohort studies cannot distinguish whether a few suffer from large LE loss or everybody loosing a little (e.g. Pope et al., 2002)
Motivation: the impact on mortality Ideally, the effects on mortality risk can be characterised as shifts in individual (population) survival curves. The impact of mortality can be expressed in two ways as a] a change [reduction] in life expectancy [∆LE], or b] a change [increase] in the probability of dying prematurely [∆RISK ] while a] is given as a difference between two integrals of survival function (1 minus b]) thorough one’s life Total benefits in population of N units is typically valued as a] N * ∆LE * VOLY b] N * ∆RISK * VSL
LE gain valuation review Johannesson and Johansson (1997) derive VOLY from WTP for a medical treatment that will increase expecting remaining length of life from 10 to 11 years when she will be 75 Soguel and van Griethuysen (2000) determined a VOLY for air pollution indirectly by asking to rank relative importance of life years lost, mild morbidity (RAD) and severe morbidity; no direct question was however asked about risk reduction or life expectancy Morris & Hammitt (2001) offered to one half of their respondents a life expectancy gain and to second one a reduction of the risk of dying and find that median WTP for the former is higher than median WTP for latter in the group that will take a vaccine in their 60 Desaigues et al. (2004; 2007) on quite small samples experimented with several WTP question variants including a) LE gain of 1,3 and 12 month, b) LE gain corresponding to ∆R=5/1000, and c) showing LE gain as well as ∆R=5/1000.
LE gain valuation: A review /2 VOLY derived from VSL assuming that VSL is the sum of discounted annual VOLY over 30 to 40 years (e.g. ExternE, 2009) Chilton et al. (2004) derive VOLY from the WTP, for the rest of their life, for 1, 3 and 6 months of extra life in normal health as well as in poor health Desaigues et al. (2007; 2011) asked for WTP for 3 and 6 month of life extension when the LE gain is not a matter of additional months of misery at the end of one’s life, but is described rather as pollution causes accelerated aging
Source: Desaigues et al. (2007; 2011)
Motivation: Which impact metric? VSL x VOLY affects the apparent merits of regulatory program that disproportionately affect people with different life expectancies (and age) – benefits of elderly may be much larger using a VSL than a VSLY approach (Hammitt 2007) Æ VOLY called as "senior death discount" (due to shorter expected lifespans of elderlies) – constant VOLY assumes that the VSL is strictly proportional to remaining LE (that is unwarranted) VOLY in use? – US EPA (1999), but only as a second alternative (Desaigues et al., 2004) – ExternE (1996-2009) but VOLY derived from VSL – ExternE (2009 on) based on Desaigues et al. (2007; 2011) – recently, the US EPA‘s SAB rejects using the VOLY approach, similarly Pearce et al. (2006; in OECD CBA) recommends to use VSL rather than VOLY approach
Our research goal • Check consistency across VSL and VOLY by doing SP study with split samples • to elicit the WTP for specified risk reductions, derive the implicit VOLY from VSL, and compare such VOLY with an estimate of the VOLY based on WTP for gains in life expectancy 1. value mortality risk reductions Æ VSL 2. value gains in life expectancy corresponding to the mortality risk reductions in 1. Æ VOLY 3. value mortality risk reductions with reminder of corresponding life expectancy gains Æ VSL* (implied VOLY*)
Revised research goal • Respondents were unable to grasp the notion of gains in life expectancy correctly in the pre-survey – people tend to think of an extension of their life by some months or years, while – the LE is an integral of the survival function thorough one’s life • Analyze the effect of reminding the LE gain corresponding to several profiles of the mortality risk reductions Æ SP study with split samples 1. Value mortality risk reductions Æ VSL 2. Value mortality risk reductions with reminder of corresponding life expectancy gains Æ VSL* • Other split sample treatments and methodological issues to study
Motivation: Contextual VSL (already studies) Subjective perception of the risk – dreaded risk, controllability, exposure, salience etc. – e.g. Alberini and Ščasný (FEEM WP 2010; under review in JRU) Mode of delivery – public program vs. private action – e.g. public premium analysed in Alberini and Ščasný (2010; 11) The cause of death – VSL may vary for specific cause which risk were reduced – e.g. Alberini and Ščasný (2011) use a choice experiments to derive VSL’s for cancer, respiratory illness and road traffic accidents related risks within VERHI-Children project
Our research goal (in the Exiopol project) • whether the cause of death on the VSL matters (all cases, cancer, cardio- and cerebro-vascular and respiratory illness) H0: cancer premium • whether environmental context matters (environmental exposures v. other) H0: VSL larger for enviro context
• whether the duration of risk reduction matters (i.e. the risk reduced permanently (i.e. sustained reductions) or during certain period (i.e. the blip)) H0: WTP for permanent risk reduction is larger than is implied by a sequence of the blips All in private good context (no public program as e.g. in VERHI project)
Choice Attributes and Their Levels Size of risk reduction
2, 3, 4, and 5 in 1000 per decade
Latency
0, 2, 5, 8 years if blip, then the risk reductions last for one decade;
Duration
Cost
if permanent, the risk reduction lasts 4 decades (if the respondent is aged 40-49), or 3 decades (if the respondent is aged 50-60) annual for the next 10 years, starting this year. The amounts: 250€, 500€, 1000€, 1800€, 3000€ (and their eq. in pounds and Czech crowns in PPP).
Choice set
Risk Communication
•
risks expressed as X in 1000 over 10 years (=X in 10,000 per year) A short probability tutorial plus a quiz whether the respondent grasped the concept (select a person with higher prob of dying)
Exper Treatment: LE gain reminder The effect of couching the risk reduction in terms of LE gains v1: risk reduction only v2: with the reminder of expected life time for that person v3: with the reminder on LE gain that is (3x) larger than v2 • …note that we did not use the expression “life expectancy” gain. We just showed current life expectancy (e.g., 79.7 years), and then the LE under each alternative. • Expected life time gain computed from life tables considering person age and risk reduction attributes (i.e. latency and the size of risk reduction). But use same for gender and countries.
Life Expectancy Gain Reminders [used in Treatment 1, version 2] Blip Latency (iin years) 0 2 3 8
dR=2 3 weeks 3 weeks 2 weeks 2 weeks
dR=3 4 weeks 4 weeks 3 weeks 3 weeks
dR=4 6 weeks 5 weeks 5 weeks 4 weeks
dR=5 7 weeks 7 weeks 6 weeks 5 weeks
Permanent Latency (iin years) 0 2 3 8
dR=2 6 weeks 5 weeks 5 weeks 4 weeks
dR=3 2.1 months 8 weeks 7 weeks 6 weeks
dR=4 2.8 months 2.5 months 2.1 months 8 weeks
dR=5 3.5 months 3.1 months 2.6 months 2.4 months
Choice set (with the reminder of expected life time)
Split Sample Treatments Treatment 1: couching of risk reduction [3] (with and without associated life expectancy gains) Treatment 2: cause of death [3] (all, cardiovascular + respiratory, cancer) Treatment 3: environmental exposures [2] (the risks being reduced are associated with environmental exposures vs. no such language) Treatment 4: repeated questioning [2] best/worst v. best/second best
Treatment 5: CAWI vs. CAPI [2] only in the Czech Republic
Survey Administration and Sampling Plan Three countries—Italy, the UK, and the Czech Republic Respondents must be – 40-60 years old – 50:50 men-women – Representative for education and income COUNTRY
MODE and PLACE
SAMPLE SIZE
Italy
CAWI (internet panel, on-line survey), 7 cities
2369
UK
CAWI (internet panel, on-line survey), 7 cities
2426
Czech Republic
CAWI (internet panel, on-line survey), 5 cities CAPI (computer assisted personal interview, at people’s homes), 5 cities + rest of the country
895 2367
The Data ITALY
UK
CZECH
Male, %
49.4
51.1
47.7
Age 40-49, %
50.3
50.0
55.3
Age 50-60, %
49.7
49.6
44.4
32,392
39,277
22,065
20.0 5.8
26.9 9.7
21.2 8.1
29.1
24.5
16.2
Hhold_income, PPP euro Own health, % (1=excellent, 2=very good) (5=poor)
Self-assessed own LE, % (longer than the average person)
Percent responses to the conjoint choice questions by treatment 1 (life expectancy gain)
Response option
1=no LE reminder
2=true LE gain
3= larger LE gain
Alt. A
35.7
24.8
27.7
Alt. B
38.4
30.1
31.9
Neither (status quo)
25.9
45.1
40.3
Random Utility Model D is the marginal utility of a unit risk reduction 'R is the risk reduction per year, G is the discount rate and L latency
Indirect utility depends on discounted flow of risk reductions and residual income; we assume that people do not discount money
D 'R e G L [1 PERM (e G 10 e G 20 AGE 40 e G 30 )] E ( y C) H
V
Eis the marginal utility of income y is income C is the payment
This last term applies only to respondents aged 4049, i.e. AGE40=1
Econometric Model and the VSL
Conditional Logit VSL computed as
VSL for cancer etc.
VSL VSL
(D / E ) 1000 (D1 D 2 ) / E *1000
Results> cause of death ITALY coeff. t. stat ALPHA1 ALPHA2 ALPHA3 BETA DELTA
6.352 0.286 0.342 -17.507 3.327
11794 -12778.0
N log L
all causes cancer cardioresp.
V
0.5864 0.0224 0.0268 -0.0003 0.0288
std. err. VSL (m€) (m€) 2.215 0.304 2.300 0.317 2.317 0.324
UK coeff. t. stat
CZ--CAWI coeff. t. stat
0.263 0.0195 0.1587 -0.0003 -0.0034
4.297 1.1772 0.39 -0.4379 2.556 -0.3529 -21.307 -0.0004 -0.334 0.078
12084 -12976.5
4470 -4735.3
VSL std. err. (m€) (m€) 0.818 0.173 0.878 0.179 1.311 0.260
5.198 -2.13 -1.721 -16.017 3.825
VSL std. err. (m€) (m€) 2.734 0.481 1.717 0.437 1.915 0.418
CZ--CAPI coeff. t. stat 0.8714 0.3071 0.2168 -0.0004 0.079
7.206 2.414 1.728 -26.404 6.924
11835 -12561.0 VSL std. err. (m€) (m€) 2.024 0.257 2.737 0.292 2.527 0.270
(D1 D 2 CANC D 3 CARDIO) 'R eG L [1 PERM (eG 10 eG 20 eG 30 )] E ( y C ) H
Results from main model. All countries.
All countries coeff. ALPHA1 BETA DELTA
no LE gain reminder
ALL
0.6088
t stat. 12.259
coeff. 1.9424
t stat. 19.858
LE reminder==2 coeff. 0.0301
t stat. 1.334
LE reminder==3 coeff. 0.2924
t stat. 4.407
-0.0003 -40.178 -0.0004 -25.594 -0.0004 -27.076 -0.0003 -22.792 0.0344
6.151
0.0709
14.66 -0.0578
-2.164 -0.0015
N
40183
13100
13594
13489
VSL (mEuro)
1.771
5.178
0.085
0.874
s.e. (mEuro)
0.125
0.229
0.062
0.178
-0.132
Results from main model. All countries.
All countries coeff. ALPHA1 BETA DELTA
no LE gain reminder
ALL
0.6088
t stat. 12.259
coeff. 1.9424
t stat. 19.858
LE reminder==2 coeff. 0.0301
t stat. 1.334
LE reminder==3 coeff. 0.2924
t stat. 4.407
-0.0003 -40.178 -0.0004 -25.594 -0.0004 -27.076 -0.0003 -22.792 0.0344
6.151
0.0709
14.66 -0.0578
-2.164 -0.0015
N
40183
13100
13594
13489
VSL (mEuro)
1.771
5.178
0.085
0.874
s.e. (mEuro)
0.125
0.229
0.062
0.178
-0.132
Results from main model mean VSL (std.err.) in million Euro no LE gain reminder
ALL Italy
UK Czech CAWI Czech CAPI ALL
LE reminder
LE reminder X3
2.273
5.869
0.415
1.027
0.264
0.504
0.265
0.391
0.878
6.172
0.195
0.175
0.468
0.125
2.183
4.295
1.096
1.406
0.354
0.366
0.776
0.535
2.425
4.295
1.495
1.924
0.215
0.366
0.519
0.343
1.771
5.178
0.085
0.874
0.125
0.229
0.062
0.178
Responses to the conjoint choice questions by treatment 2 (environmental exposure) Response option
1=mention of 0=no mention environmental exposures
Alt. A
28.6
30.1
Alt. B
32.7
34.2
Neither (status quo)
38.7
35.8
Results: environmental context. All countries
All countries
no envir. context
envir. context
coeff.
t stat.
coeff.
t stat.
0.408
6.35
0.8084
11.139
BETA
-0.0003
-27.827
-0.0004
-28.965
DELTA
0.0143
1.584
0.0485
6.941
N
20170
20013
VSL (mEuro)
1.213
2.306
s.e. (mEuro)
0.169
0.177
ALPHA1
The VSL difference is significant at 1% level
Results: environmental context, mean VSL (s.e.)
no enviro context
ALL Italy Czech CAWI Czech CAPI ALL (CZE,ITA,UK)
with enviro context
2.273
1.918
2.642
0.264
0.38
0.368
2.183
2.201
2.173
0.354
0.453
0.557
2.425
1.687
3.066
0.215
0.332
0.284
1.771
1.213
2.306
0.125
0.169
0.177
Results Responses pass the scope test—except for first version of the life expectancy treatment (v2)
Responses consistent with the economic paradigm – BETA – MU of income positive and significant – ALPHA – MU of the risk reduction is positive and significant VSL for both LE gains are much lower and not different one from the other Low discount rate—but suggests that people do not value future risk reduction as much, even if permanent
But… that model doesn’t cater to the possibility that people might care more for sustained risk reductions than for a sequence of blips. Revised indirect utility functions Æ allows different marginal utility of risk reductions for blip (D1) and sustained risk reductions (D2)
V
'R e
G L
[D1 BLIP D 2 PERM (e
G 10
e
G 20
e
G 30
E ( y C) H … compare to
D 'R e G L [1 PERM (e G 10 e G 20 AGE 40 e G 30 )] E ( y C) H
V
)]
Results: different MU of dR for blip v. permanent Only sample with no mention of LE gains (treatment1, v1) ALL coeff t stat alpha1 alpha2 beta delta
blip permanent
V
ITALY coeff t stat
UK coeff t stat
Czech Rep coeff t stat
1.9752
21.113
1.8917
11.027
2.1889
12.960
1.9439
12.673
1.2113
10.18
1.4836
6.212
1.4789
6.504
0.8914
5.289
-0.0004
-26.395
-0.0003
-12.277
-0.0004
-13.637
-0.005
-19.286
0.0279
4.063
0.0401
3.205
0.0268
0.011
0.0219
1.763
VSL (mill. Euro) 5.029 3.084
VSL (mill. Euro) 5.802 4.550
VSL (mill. Euro) 5.978 4.039
VSL (mill. Euro) 4.128 1.839
'R e G L [D1 BLIP D 2 PERM (e G 10 e G 20 e G 30 )]
E ( y C) H
Results: …with an alternative-specific intercept for sustained risk reductions (D2PERMANENT) Only sample with no mention of LE gains (treatment1, v1) ALL coeff alpha1 alpha2 beta delta
VSL based on alpha1
ITALY
t stat
1.9818 0.2161 -0.0004 0.0962
19.570 5.525 -25.600 12.051
coeff
t stat
1.9436 0.2936 -0.0003 0.0948
10.360 4.024 -12.140 6.243
UK coeff
Czech Republic
t stat
2.2418 0.2671 -0.0004 0.1083
12.005 3.527 -12.255 7.039
coeff
t stat
1.8921 0.1149 -0.0004 0.0851
11.878 1.984 -18.602 7.923
VSL in mill. € (s.e.)
VSL in mill. € (s.e.)
VSL in mill. € (s.e.)
VSL in mill. € (s.e.)
5.297 (0.240)
6.128 (0.546)
6.402 (0.502)
4.262 (0.304)
D1 'R e G L [1 PERM (e G 10 e G 20 AGE 40 e G 30 )] PERM D 2 E ( y C ) H
V
Summary and conclusions SP study that looked at permanent v. blip within the context of reminding people of the LE gains implied by mortality risk reductions Results satisfy scope test and economic paradigm
Reminding people of the LE gains implied by a mortality risk reduction reduces their WTP, and hence the VSL VSL under environmental context not significantly different for country samples People are willing to pay more for a sustained risk reduction than for a sequence of blips, but WTP is only slightly larger
All results so far being presented based on using only the response to the first conjoint choice question for each A-B-status quo choice task. Analysis based on repeated responds (best/best, best/worst) from the same survey will come tomorrow
VSL for CAWI vs. CAPI in Czech Republic
Sub-sample
CAPI
CAWI
ALL
2.48
2.25
Cities (5)
2.55
2.25
Prague
4.02
3.85
Cities (4)
1.79
1.37
age40-49
2.33
2.44
age50-60
2.76
1.94
Contacts: Dr. Milan Ščasný Charles University Prague, Environment Center [email protected] Dr. Anna Alberini Fondazione Eni Enrico Mattei & University of Maryland, USA [email protected]
