These days, anyone who’s anyone has got an election forecast. But we think we may have discovered the first widespread exercise in election forecasting.
In October 1924, the Daily Mirror, in conjunction with the Sunday Pictorial, ran a competition for its readers to forecast that month’s general election. Entrants had to predict the correct number of MPs for each of the three main parties, along with a figure for the others (“representing any independent parties, such as self-described Communists and Prohibitionists”).
The rules were precise: entries had to be “written in ink” on a coupon from the paper (which should be “neatly cut out”).The prize was a whopping £1,000, around £50,000 in today’s money. The article ran next to one on young people and “purity”, with the sub-head: “Young People Hardly Know What Is Meant, Says Woman”. Some things don’t change.
In October 1924, the Daily Mirror ran a competition…
Announcing the result in November, the paper claimed it had received “many millions” of entries. The winner – the only entry to get it spot on – was what the paper called a “twenty-four-year-old village girl”. Kathleen Cotton, of Gnosall, Staffordshire, was the daughter of the local plumber and decorator, and it was, apparently, the first competition she had ever participated in. “Village Maid Only Reader to Forecast Exact Result” (as the Mirror put it).
In an interview with the paper, “the golden-haired ‘£1,000 girl’” explained her methodology. The first thing to do was to listen to those who knew about the subject, and base your predictions on what they said – what would today be called anchoring.
In particular, she had taken note of a prediction made by Sir Patrick Hastings about the fate of the Liberal Party, and based her predictions on that. However, it turned out that the second thing to do was to not listen very carefully to those who knew about the subject, because she revealed she had misunderstood what Hastings had said.
He had predicted that the Liberals would win 80 seats. She had misread that, for 50, and thus kept all of her Liberal predictions below 50. Had she anchored on his actual prediction, she wouldn’t have won.
The third thing to do – entirely within the Mirror’s rules, but kind of breaking the spirit of forecasting in its modern sense – was to send in multiple entries. She had sent in a total of 15 separate coupons (presumably involving buying 15 separate editions of the paper, although this wasn’t made clear).
The first thing to do was to listen to those who knew about the subject, the second thing to do was to not listen very carefully.
Whilst the Mirror reported receiving “many millions of entries”, they also referred to her beating “hundreds of thousands of readers”, which implies that lots of people were doing this – or perhaps that the Mirror was also somewhat inflating the popularity of its own competition (perish the thought).
And her last tip was to not cluster your predictions. “Judging by my own experience”, she said, “I should say that the best way to forecast anything is to go, by degrees, from one extreme to the other”. Her predictions for the Conservatives had ranged from a 119 seats up to 415. Philip Tetlock would recognise this.
The first thing Kathleen Cotton intended to do with her £1,000 was go for a “week-end’s shopping in London”; the second “was to purchase a motor-car”.
It must be safe to assume, alas, that the Mirror’s ‘many millions’ of coupons have now been long destroyed, else they would provide a lovely early test of the wisdom of crowds thesis.
One curious thing, though. Miss Cotton’s winning prediction was Conservative: 415; Liberal 43; Socialists 151; and other parties: 6. This is not how the election is usually reported today. Rallings and Thrasher’s British Electoral Facts gives 412, 40, 151, 12 (as does Wikipedia). Butler and Butler’s British Political Facts gives 419, 40, 151, 5.
But as the Mirror made clear: “The Editor’s decision shall be final and binding in all competitions”.
Matthew Bailey and Philip Cowley are both from the Centre for British Politics at the University of Nottingham,