\magnification=1200
\baselineskip=20pt
\nopagenumbers
\font\big=cmr12 scaled \magstep2

\centerline{\bf STANFORD UNIVERSITY}
\centerline{\bf DEPARTMENT OF STATISTICS}
\centerline{\big DEPARTMENTAL SEMINAR}
\bigskip

\baselineskip=12pt
\centerline{4:15 p.m., Tuesday, November 27, 2001}
\centerline{Sequoia Hall Rm. 200}
\centerline{(Cookies at 3:45 in 1st Floor Lounge)}

\bigskip
\baselineskip=15pt
\centerline{\sl Simon Jackman}
\centerline{\sl Dept of Political Science}
\centerline{\sl Stanford University}
\bigskip
\centerline{\bf State-Level Election Forecasting during Election 2000 via Dynamic Hierarchical Modeling}

\bigskip

During the 2000 presidential election campaign, we developed a forecasting model for the 51
state-level election outcomes and the aggregate Electoral College result, relying heavily on
Bayesian principles.  The Bayesian approach provides a convenient and flexible framework for
integrating a large amount of proprietary state-level polling (by Knowledge Networks) with other
sources of relevant information, such as published state-level poll results, previous
presidential elections, and national-level polls.  This approach is easily generalized to deal
with dynamics, making it a feasible methodology for "tracking" candidate support at the
state-level over election campaigns.  We discuss the performance of this model in the 2000
election, and suggest how the methodology can be deployed in other electoral settings.

(This work is joint with Doug Rivers, of the PoliSci Dept and also Knowledge Networks, Inc.)

\bye