Context, latency and the value of preventing a statistical cancer fatality

Abstract

This thesis contributes to the state of understanding about the value of latent health and fatality risk reductions, focussing on the effects of context and latency on the Value of Preventing a Statistical Cancer Fatality (VSLCAN) relative to road accident fatalities. The conceptual, methodological and empirical contributions are derived from two stated preference studies. The studies are designed to explore how the VSLCAN is driven by the context effect, which includes dread of the cause ‘cancer’ and the effects of illness prior to fatality; and the latency (delay) effect which depends upon time preferences and risk preferences. Study 1 develops a Risk-Risk survey protocol, and the resulting central tendency and regression analysis verify that the context of cancer increases the VSL and that latency decreases it. The relativity between VSLCAN and the road accident VSL is then summarised into a simple relationship where the offsetting influences of context and latency are parameterised. This novel tool has the potential to enhance the comparability and evaluation of a wide range of existing and future VSL studies involving context and latency effects through the elicitation of key underlying parameters such as the context premium and effective discount rate. As such it represents a significant methodological contribution. Study 2 focusses directly on two aspects of the latency effect. These relate to risk and time preferences, explored in Studies 2a and 2b respectively. Delayed outcomes are inherently risky, so the exploration of latent outcomes requires controlling for risk preferences. Study 2a develops a theoretical and empirical framework for eliciting risk aversion proxies in the domain of health, which have not previously been fully developed in the literature. The method extends the classic Holt-Laury risk preference elicitation framework into a new domain- health risks- and the method is implemented successfully in Study 2. This chapter therefore makes both conceptual and methodological contributions through clarifying the utility theoretic basis of a health risk aversion measure and then developing a way to elicit such a measure in surveys. Study 2b uses the novel VSLCAN:VSL relationship developed in Study 1 to elicit exponential discount rates from Risk-Risk data comparing latent cancer and road accident risks. Regression analysis performed on these rates on a sample and individual level, provides strong evidence to suggest that a non-standard (sub-additive) discounting model is the most descriptively accurate discounting assumption for this sample. It provides the first evidence regarding sub-additive discounting in the domain of health and fatality risk.