USING ELICITED CHOICE EXPECTATIONS TO PREDICT CHOICE BEHAVIOR
IN SPECIFIED SCENARIOS

Charles F. Manski

Revealed preference analysis and hypothetical choice analysis
offer complementary approaches to prediction of choice behavior. 
Revealed preference analysis yields predictions by combining
observations of realized choices with assumptions about
underlying decision processes.  Hypothetical choice analysis
yields predictions by combining observations of elicited choice
expectations with assumptions about underlying processes of
expectations formation and communication.  A longstanding concern
of research in economic theory and in econometrics has been to
illuminate the logic of the revealed preference approach to
prediction of choice behavior.  The logic of hypothetical choice
analysis has not received comparable attention.  As a
consequence, the practice of hypothetical choice analysis has
suffered from a lack of rigor, indeed from an absence of
coherency.

This paper continues the study of the logic of hypothetical
choice analysis begun in Manski (JASA, 1990).  There I examined
the predictive power of intentions data (responses to questions
asking for unconditional point predictions of future  choices)
under the "best-case" hypothesis that individuals have rational
expectations and that their stated intentions are best point
predictions of their future choice behavior.  Here I am concerned
with the use of probabilistic expectations data to predict
behavior in scenarios posed by the researcher.  To illustrate
these distinctions, consider these three questions eliciting
fertility expectations:

Looking ahead, do you expect to have any (more) children?

Looking ahead, what is the percent chance that you will have any
(more) children?

Imagine that the government were to enact a child-allowance
program providing families with fifty dollars per month for each
dependent child. Assuming that this program were in operation,
what is the percent chance that you would have any (more)
children?

The first question calls for a statement of fertility intentions. 
The second question calls for an unconditional probabilistic
prediction of future fertility. The third question specifies a
scenario and calls for a probabilistic prediction of future
fertility in that scenario.

After laying out basic concepts of probabilistic choice analysis,
I perform a best-case analysis in the spirit of my earlier work. 
I then discuss relevant logical and cognitive considerations.