Development of a Theory of Values and Their Measurement
A theory of values and their measurement will be developed on the basis of two distinctions among types of values: protected values vs. well-behaved values, and fundamental vs. proxy values. Protected values, in contrast to well behaved values, resist tradeoffs with other values and concern means rather than ends. For example, some people think that destruction of species by logging should be stopped at any cost. Proxy values are stan-ins for fundamental values, to which they are related through beliefs (often uncertain). For example, percent of children vaccinated is a proxy for disease prevention to which it is related by beliefs about vaccine effectiveness. These distinctions can help us understand inconsistency involve insensitivity to quantity (e.g., insensitivity to range in assigning relative weights to attributes, quantity insensitivity in contingent valuation), and nonconsequential principles that lead to different valuations of identical outcomes achieved by different means (e.g., use of cost rather than benefit in contingent valuation, omission bias). Hypotheses regarding the sources of these inconsistencies will be tested by examining the effects of: fundamental vs. proxy values; information relating to proxy fundamental values; manipulations of the measurement task to bring subjects' approach in line with its purpose; and combinations of these manipulations. It will also be determined whether increased consistency within each of two measures will increase agreement between them.
|Pennsylvania, University of|
Cost to Funding Agency:
|Project Status Reports:
April, 1997: The ultimate purpose of this research is to design improved measures of the utility of outcomes, such as the effects of policies on the environment or on individual well-being (in the form of time or monetary expenditures or health effects). All current measures - contingent valuation, standard gambles, direct ratings, time tradeoffs, person tradeoffs, and multi-attribute judgments - have been criticized for being internally inconsistent, insensitive to relevant factors, or sensitive to irrelevant factors.
In utility measurement tasks, subjects judge the utility of objects relative to some zero value. For example, subjects may judge the disutility of losing a finger or a hand relative to the zero point of losing nothing. In some judgments, subjects produce the object rather than the relative judgment. So, they might indicate the amount of money that has half of, or all of, the utility of losing a finger (as in contingent valuation).
Judgments can be indirect in terms of hypothetical decisions such as buying or gambling. Indirect judgments are distorted by heuristics and intuitions concerned with the context of the decision rather than the values of interest: values seem protected from tradeoffs when deontological moral rules (rules prohibiting certain actions) are brought to bear; subjects confuse means with ends, and cost with value; subjects are insensitive to quantity; and subjects confuse quantity measures with each other (e.g., number of lives saved with proportion of lives saved).
Direct judgments avoid many of these problems, but they raise other problems, in particular, those concerned with the scaling of utility. For example, subjects may say that losing a finger has half the (dis)utility of losing a hand but losing a hand has ten times the (dis)utility of losing a finger. This may result from excessive attention to the subject of the comparison (finger or hand, respectively).
Our current research is directed at solutions to these problems of direct judgment, as well as continued investigation of the distorting effects of indirect methods. One type of solution involves the use of tasks designed to minimize scaling distortions, plus consistency checks and the opportunity to revise initial judgments. For example, the problem of inconsistent proportions may be avoided by using neither finger nor hand as the standard and asking for utility judgments of both simultaneously. The resulting judgments may be checked for conformity to other criteria such as transitivity (finger/arm ratio should be product of finger/hand and hand/arm). We find that most subjects can correct their judgments based on such criteria without feeling that their judgments are no longer honest representations of their views.