Listen to the dawn of data journalism: Univac, the first Nate Silver

In one of my data journalism presentations I look back at the history of data journalism, and one of the key dates is 4 November 1952: the first US election when a TV network, CBS, used a computer to analyse and predict the election results.

CBS used the room-sized, vacuum-tube powered Univac.  The idea came from Remington Rand, the makers of the Univac, because sales weren’t as strong as they wanted. In a creative bit of marketing, they approached CBS to use the computer to help analyse the election results. Of course, CBS also saw a marketing opportunity and mentioned Univac in their election coverage ads. Last night, in a lovely bit of luck, I heard one of those ads.

I often listen to old radio dramas while I’m making dinner and, much to my surprise, last night when I was listening to the 13 October 1952 episode of the western radio drama Gunsmoke on Old Valve Radio, I heard CBS run an advert touting how “Univac, the electric brain” would be assisting Edward Murrow and his team on election night coverage the following week. My jaw just dropped to hear this message announcing the dawn of computer-assisted reporting, as it was called in the US, 70 years before the term data journalism came into vogue. You can hear the ad yourself on Podbay.fm, and, if old western radio drama isn’t your cup of tea, fast forward to 13:43 to hear about how the CBS election team would be backed up by “Univac, the electric brain”.

For a lot of data journalists, the CBS-Univac partnership is a famous and well known bit of history, but if you aren’t familiar with the rest of the story, it is fascinating. Both Remington Rand, makers of the Univac, and CBS saw value in the arrangement, as Wired explained when they looked back at the day in 2010.

The eight-ton, walk-in computer was the size of a one-car garage and accessed by hinged metal doors. Univacs cost about $1 million apiece, the equivalent of more than $8 million in today’s money. The computer had thousands of vacuum tubes, which processed a then-astounding 10,000 operations per second. (Today’s top supercomputers are measured in petaflops, or quadrillions of operations per second.)

Remington Rand (now Unisys) approached CBS News in the summer of 1952 with the idea of using Univac to project the election returns. News chief Sig Mickelson and anchor Walter Cronkite were skeptical, but thought it might speed up the analysis somewhat and at least be entertaining to use an “electronic brain.”

They had no idea how quickly it would speed up analysis, and early on the evening with only about 5 percent, or 3.3 m of the total 61 m votes, counted Univac had a prediction. The Computer History Museum has the printout of a prediction the Univac team sent to CBS via teletype at 830 pm. “It’s awfully early, but I’ll go out on a limb.”

However, just as traditional political pundits heaped scorn on stats wizard Nate Silver in 2012, CBS’s editors were not willing to join the Univac team out on that limb. On air, CBS hedged:

In another story about the election by Ars Technica, we learn that the Univac team figured out that they had entered the New York results incorrectly, overstating Stevenson’s votes by a factor or 10. They ran the numbers again, but they got the same result. Univac predicted that Republican Dwight Eisenhower would win the Electoral College 438 to Stevenson’s 93 votes, and the computer set the odds of an Eisenhower win at 100-1.

As the night went on, Eisenhower gained momentum. The final vote was 442 to 89 to Eisenhower, and Univac’s early prediction was off by only one percent.

While I’ve known about this story for a while, it was great to hear the advert of CBS advertising how it was going to use an “electric brain”. I also learned something new about Univac. It was programmed by computer pioneer Grace Hopper. Her team fed the computer with statistics from previous elections, and she actually wrote the code that allowed Univac to make its prediction. Sadly, her contribution was not mentioned in reports at the time.