In 1989 the Exxon Valdez ran aground and spilled 11 million gallons of
oil in Alaska. Great controversy has surrounded the assessment of mon
etary damages for this oil spill. The source of this controversy has b
een the use of hypothetical contingent valuation surveys to assess dam
ages to individual households and the aggregation of these survey valu
es over all households in the United States. The assessment prepared f
or the state of Alaska is reported to have cost several million dollar
s to undertake. We show that considerable economy can be effected in t
he elicitation of such survey responses. Our approach is to use a chea
per ''convenience sample'' to generate a statistical model of behavior
al responses to the valuation questions and then to use sample average
s of population characteristics to predict the behavior of the populat
ion. Thus we use a model of the behavior of students to predict the be
havior of all of the adult citizens of the United States. We find that
we can obtain essentially the same damage estimates as a more expensi
ve ''probability sample'' of the entire nation. (C) 1996 Academic Pres
s, Inc.