The Electoral College Simulator models the US Presidential election using current state polling data (which can be changed if you think the polls are wrong!). The probabilities of Bush and Kerry winning each state are calculated using the poll data and normal statistical processes (Gaussian distribution, etc.). These probabilities determine which candidate wins each state's Electoral College votes in a given simulated election, yielding a specific number of electoral votes for each candidate. After running a large number simulated elections, the distribution of electoral votes for a candidate is plotted as above (which is a sample; click "Simulate" for a current plot).

How do I start a simulation?
Click the "Simulate" button at the top of the right column.

What do the "Parameters" change (upper right)?
Number of simulated elections - sets the number of elections that the Electoral College Simulator processes to produce the plot and statistics.
Bin width, max/min electoral votes, and max percent in plot - 'Bin width' is the number of Electoral College votes in each histogram bin; 'max/min electoral votes' and 'max percent' determine the bounds of plot.
Kerry/Bush offset - these parameters independently add a fixed percentage to one candidate's state poll data. It does not subtract a corresponding amount from the other candidate. This feature is useful if you think that one candidate is consistently under- or over-polling across the country, for example, because voter turnout or new voter registration is not accurately reflected in the polls.
Poll error multiplier - this parameter multiples the poll margin of error by a fixed amount, which is useful if you think that the polls are less accurate than claimed.
"Anti-correlation" parameter - this parameter attempts to model how correlated Bush votes and Kerry votes are.

Questions/Comments?
Email me (Andrew Howard), or visit my home page if you like.

visits to http://frank.harvard.edu/~howard/election/: