1998 Congressional Election Predictor and 106th Congress Simulation
About this Simulation
Purpose of this simulation
How does it work?
How accurate is it?
Credits
Purpose of this simulation
This simulation is meant to be entertaining and informative for those who are interested in American politics.
While it is intended that it be as accurate as possible, given the data available, don't go and bet anything on
its predictions as they are in no way guaranteed.
How does it work?
The assumption is made that people will vote the same way they did the last election, except for the changes specified
by the user which are applied to each race. The outcome of each Senate and House race is predicted individually.
Next, the number of winners for each party is totaled to provide party counts for both the House and Senate.
The calculation is fairly simple. The votes received by the Democratic candidate and the Republican candidate in
the previous election are totaled. The percentage of that total received by a party is used as a base percentage
for that party in the current election. These percentages are adjusted based on the parameters specified: change
in the support for incumbents, change in the support for the parties, and the additional adjustments for when the
seat is open or the incumbent is a freshman. Whichever party has the higher adjusted percent for a district is
judged to be the winner of that district for the current election.
Importantly, this model does not account for changes in the election results due to voter turnout, the merit of
the candidates, nor the effectiveness of their campaigns. The assumption is made that across the country, these
effects tend to average out.
For simplicity, it is assumed that the one independent incumbent, Representative Bernard Sanders of Vermont will
be reelected. Also, Senators Richard Shelby of Alabama and Ben Nighthorse Campbell of Colorado, who have switched
to the Republican Party since their last election, are assumed to take a penalty equal to the percentage lost by
a party vacating a seat (the open seat adjustment).
How accurate is this simulation?
With the parameters set appropriately, it correctly predicts the winner in approximately 87% of the House races
in the 1994 and 1996 elections. This is not too bad considering it has to cope with a dramatic turn against Democratic
incumbents in the 1994 election, the strong shift towards Republican candidates in several Southern states over
the last several elections and the redistricting in Georgia, Louisiana, Florida, Kentucky, Maine, and Texas. With
no such events foreseen for the 1998 election, it should be even more accurate. No comparative results are available
for the Senate since the earliest data used is from the 1992 election and only this year will those Senators be
voted on again. The model used is identical to the one for the House.
In general, a model that simple predicts that all incumbents will be re-elected will be fairly accurate. Hopefully,
the model used here bests that one. In the end, it doesn't matter if it predicts the outcome of each individual
race correctly, as long as its missed predictions of Republican and Democrat winners tend to balance out. It's
the averages that matter.
Credits
The principal source for the election data used is the page of Historical
Election Results and 1998
Congressional Candidates from the Office of the Clerk, U.S. House of
Representatives. Results for special elections comes from various Secretary of State web sites or, when not available,
National Journal's Almanac of American Politics
The Congressional roll call votes come from the House
and Senate web sites, with additional descriptions
from the Library of Congress' Thomas site.
Information on the status of incumbents come from the Office
of the Clerk, U.S. House of Representatives; Roll
Call Magazine; and Project Vote Smart. A thank you goes to Heather
Holdridge, head of Project Vote Smart's Reporter's Resource Center and my former colleague, for her help with this.
The Java slider class used in this program originated from the Core
Web Programming site.
If this program does not appear to work correctly in your browser, I'd like to hear about it. Please identify the
type of browser you are using, its version number, and the operating system you are running. Send this information
and any other comments or suggestions about this simulation to the author at sl@scottlangley.com
Enjoy!
Scott Langley
© 1998 Scott Langley