• Please review our updated Terms and Rules here

UNIVAC 1952 election prediction

M

mechaniputer

Experienced Member
Joined
May 23, 2015
Messages
59
In 1952, UNIVAC famously predicted the outcome of the presidential election based on a partial vote tally. I am trying to figure out exactly what algorithm was used. I know that the algorithm produced an overall prediction of the electoral college votes as well as a value ((percentage?) representing the certainty of that prediction.

I know that they "trained" it based on results from past years. However I don't even know the type of data they used. Was it just per-state totals? A record of the counting process over hours or days? Something else?

In 1956 all major networks included computer prediction for election coverage. Maybe there's more info about that. What machines was that done with?

Does anybody have more info or an educated guess?
 
More info. CBS was the network airing UNIVAC's prediction, but apparently NBC had a smaller computer of their own called Monrobot III, made by Monroe Systems for Business. I can hardly find anything about that machine. But it likewise predicted the outcome correctly.
 
Chuck(G) said:
The Monrobot III was scarcely more than a calculator
Click to expand...

Thanks! Well, it did store programs on the drum, and it did have conditionals. The memory was quite small though. It had a "modification" instruction to change the operands of instructions, which if I understand correctly, could be used for implementing a stack or array. For example they claim that for printing out a sequence of 50 values from memory, only one MOD and one PRINT are required (in a loop, possibly with other steps but this is unclear). The instruction set looks general enough to have been hypothetically useful with much larger memories, without requiring the size of the input data to be known in advance. I'd call that a computer, albeit with a lacking memory capacity.

I'd still like to figure out what UNIVAC was doing that night.
 
It was more a publicity stunt from Remington Rand than a scientific breakthrough. Dig some and you'll see that the predictions were being made on the basis of so-called "bellweather" earlier election results; mostly primaries. Max Woodbury designed the algorithms used. Consider the computing power of UNIVAC I and you can readily see that the same result could be had using a contemporaneous desktop mechanical "adding machine", such as those that Monroe, Victor, etc. had already placed in corporate accounting offices and banks. Your reference to the Monroe "Monrobot III" further reinforces that idea.

The predictions weren't all that earthshaking if you look at the political climate. In "Mr. Citizen" you can see where Truman was an SOB who refused to endorse Stevenson. There was no incumbent candidate; the Korean war was very unpopular. Ike had his wartime record as Supreme Allied Commander; it's very difficult to see how Eisenhower could not have won. This was also the first US presidential election to be covered on television.

And the pundits were very often wrong. Consider the Dewey-Truman election four years earlier, where the Chicago Tribune actually published a lede claiming Dewey's victory.

Finally, you have to understand the state of advertising and public relations. Postwar PR and advertising was really just getting started with the idea that not only could you appeal to consumer values but that you could create them.
 
Last edited:
Chuck(G) said:
. . .. Ike had his wartime record as Supreme Allied Commander; it's very difficult to see how Eisenhower could not have won. This was also the first US presidential election to be covered on television.
Click to expand...

I was 12 years old at the time of the 1952 presidential election, and had an "I Like Ike" T-shirt. Computers were more or less a novelty back then, as Art Baker, on the "You asked for it" tv show had a fellow with an abacus go up against some computer and the abacus won. Stevenson lost because no one knew who he was. Same with his running mate, John Sparkman. Almost identical circumstances in the 1956 election and the country was running "smooth". When people mentioned computers back then, they were mostly referring to IBM.
 
My parents would refer to computers as "IBM Machines"; for that matter, an 082 sorter would be treated in the popular press as a "computer".
We wouldn't consider a 407 accounting machine to be a computer, being plugboard programmed and all, but serious computation was done with those, particularly when coupled to a 519 punch.
 
Chuck(G) said:
It was more a publicity stunt from Remington Rand than a scientific breakthrough.
Click to expand...

No doubt. I'm just curious about what they were able to do on the machine with regards to the problem, even if it wasn't particularly novel.
 
Chuck(G) said:
My parents would refer to computers as "IBM Machines"; for that matter, an 082 sorter would be treated in the popular press as a "computer".
We wouldn't consider a 407 accounting machine to be a computer, being plugboard programmed and all, but serious computation was done with those, particularly when coupled to a 519 punch.
Click to expand...

1957-58 I was in the Navy as an airman apprentice (E-2) and was assigned to the Supply & Fiscal Department at a Naval Air Station in south Texas. All issues and receipts were done by hand and on Kardex files. The main Supply Center, at another location, was transitioning to IBM key punch. The civilian ladies in my unit were fighting for the local 'key punch operator' positions as the GS scale would be GS-5, and the highest GS grade in the department at that time was the GS-7 boss. Those IBM cards changed everything concerning aircraft parts and accountability. IBM ruled in those days.
 
You must log in or register to reply here.
Back
Top