Intervention (Hidden)
The project ANR-20-CE36-0010-01 (VHEALTH hereafter) aims at providing new tools to assess monetary values for improved health and longevity. The most common practice to approximate the monetary value of a gain in health or longevity (or both) is to combine preference-based measures of the severity and duration of illness, often expressed in quality-adjusted life years (QALYs), with monetary estimates of the VSL (e.g. Mason et al., 2009). The QALY metric is widely used in medical and health economics. It assumes that health is the sum of the time spent in each health state weighted by an index of health quality, where a weight of one corresponds to full health and a weight of zero to a health state as bad as dead. The guidelines for regulatory impact analysis of the U.S. Department of Health and Human Services (2016) recommend that an "approach [for valuing morbidity] is to calculate a constant value based on expected QALYs [...] by dividing the VSL by future QALYs". Dividing VSL by future QALYs implicitly assumes that the tradeoff between wealth and the probability of surviving the current period is proportional to future QALYs. However, this assumption is inconsistent with empirical evidence on how the VSL varies with age (Aldy and Viscusi, 2007; Krupnick, 2007). The approach implies that the value for improved health or longevity is constant, which violates standard assumptions of mortality risk valuation (e.g., VSL varying with health and longevity; Hammitt, 2013).
In previous work (Herrera-Araujo et. al, 2020) we explore the validity of the widely imple- mented practice of dividing VSL by future QALYs to approximate the monetary value of a health improvement. Our theoretical setting allows us to derive empirical tractable upper and lower bounds for the value for improved health and longevity. We find that dividing VSL by future QALYs corresponds to an upper bound of the WTP for improving health, which implies that dividing VSL by future QALYs overstates the monetary value of health gain. We also identify a lower bound not previously used in policy applications but appealing given its conservative nature. Our research aims at empirically identifying the upper and lower bounds for the value of improved health and longevity along with VSL estimates in France
The project’s second objective aims at providing empirical support for an innovative method for estimating the VSL through SP: the use of non-marginal risk reductions. Most papers ask respondents to value small risk reductions to elicit VSL. These have the advantage that under conventional theoretical assumptions, WTP should be nearly proportional to the proposed risk reduction for small risk reductions.
For example, if WTP for a 1 in 100,000 change in the annual risk of premature death equals 30 euros, then the WTP for a 2 in 100,000 change should be close to 60 euros. This result provides a useful scope test: responses should be consistent with the hypothesis that WTP is nearly proportional to the risk reduction. Passing a scope test is considered a necessary condition for an SP study to be of good quality. In practice, stated-preference studies pervasively suffer from a lack of scope sensitivity (Hammitt and Graham, 1999). This is called "scope insensitivity". There are at least two reasons for this. First, scope insensitivity may result from the diminishing rate at which individual’s trade income for risk reduction as the risk change increases (i.e., the curvature of the utility function; Corso et al. 2001); and second, it may result from respondents’ lack of understanding of the "good" being valued.
In designing VSL elicitation questions, there is a trade-off between distortions due to the curvature of the utility function and distortions due to respondents’ limited comprehension of the risk reduction. For small risk reductions (say, a 1 in 100000 change in the annual risk of premature death), the curvature of the utility function -the rate of change in the rate at which individual’s trade income for risk reductions- is unimportant, but the respondents’ cognitive cost of evaluating such small changes is high. For larger risk reductions (e.g., a 1 in 100 change in the annual risk of premature death), the curvature becomes more relevant and respondents should have a better understanding of the size of the risk change they are valuing.