The first computer prediction of a U.S. Presidential election was by a UNIVAC I computer used by CBS News on election night 1952. At 9:15 pm, with only 3.4 million of the eventual 61.8 million popular votes counted, the UNIVAC predicted 43 states and 438 electoral college votes for Dwight Eisenhower, and 5 states and 93 electoral votes for Adlai Stevenson. The actual results were 39 states and 442 electoral votes for Eisenhower. The popular vote was also within 3.5%. It proved to be a success for both UNIVAC and CBS.

What inputs and algorithms were used by the UNIVAC to predict the election? Was this merely a ratio of the votes counted at the time, or was there a more sophisticated mathematical model?

UNIVAC prediction of the 1952 election

  • 2
    There's a wide range of potential statistical algorithms, from using the outcome of the last elections as a baseline, to more finer grained methods that identify important voting districts and weight them. It's probably impossible to say which methods they used without seeing the source code, but as computer usage was quite maths heavy at that time, it's virtually sure they used as much statistics as the UNIVAC was able to handle.
    – dirkt
    Commented Nov 3, 2020 at 4:44

1 Answer 1


Seems like it was designed by Grace Hopper / Remington Rand:

Navy mathematician Grace Murray Hopper worked for Remington Rand with a team of programmers who input voting statistics from earlier elections and wrote code to allow the calculator to predict the result based on previous races. Technicians would use Unityper machines to encode the data onto paper tape to feed into UNIVAC.


  • 3
    I don't suppose there's any chance at all it was a publicity stunt cooked up by Remington Rand and CBS?
    – davidbak
    Commented Nov 5, 2020 at 4:21
  • 1
    It was, but they failed. CBS, famously, refused to read it because it was right. Commented Nov 5, 2020 at 20:04

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .