When we talk about Nate Silver to observers of American politics, many of them hold their noses. But the former star of “data politics” was not always so reviled. In 2008, while he was still writing under the pseudonym “Poblano”, this American baseball championship expert decided to apply his sports statistical prediction model to the presidential election. The result is resounding: it successfully predicts the victory of Barack Obama and gives correct results in 49 of the 50 American states. Four years later, he made a homerun – 50 out of 50 – while many specialists announced a very close vote between Obama and Mitt Romney. His blog, entitled “FiveThirtyEight.com” (in reference to the 538 electors responsible for electing the President of the United States), was then hosted by the New York Times. And it’s an audience hit.
Silver has never published exactly how his algorithm works. We do know, however, that it aggregates all the polls carried out during the campaign, which it weights according to their recency, the size of the samples and the partisan biases of certain institutes. The model has been emulated: since 2008, many media outlets have set up their own survey aggregators like The Economist, Politico or even more recently Parisian. Everyone is banking on the fact that opinion polls provide enough information to announce the chances of victory for each candidate, even if this has not always been the case in the past.
In 2016, for example, most of these models failed to announce the victory of Donald Trump. Silver, for his part, points out that he still gave the real estate mogul a 29% chance of winning – almost one chance in three, much more than most of his competitors. But whatever. This is a first setback for the blogger and his enemies are quick to denounce this type of model, starting with Allan Lichtman, another star of electoral predictions. The emblematic political scientist, professor of history emeritus at the American University in Washington, is one of the figures of what we call the “fundamentalists”. Proponents of these methods – often academics – prefer to base their predictions on empirical elements such as the state of the economy, the CVs of the two candidates or even their success during the debates. Lichtman, for example, has developed a system that includes 13 such elements, to which he assigns a “true” or “false” value. If the outgoing party has fewer than six false answers, then it wins the election. So far, the model is unstoppable: since 1981, the political scientist has systematically announced the winner of the American presidential election, including in 2016. However, he imagined that Donald Trump would also win the electoral college (he ultimately lost it with almost three million votes).
Today, in the American media, the methods of Silver and Lichtman are systematically pitted against each other. Each in their own way, the two men are “prophets” of the American election. Here they are, heroes of a real cultural war. On one side, a libertarian lone wolf, fond of poker and gambling, on the other a university professor close to the Washington elites. The two men even engage in a virulent battle via tweets. The election of November 5 will be their justice of the peace.
Predictive bets?
This year, a new model is enjoying wild success – but is it really one? On online betting sites, many Internet users around the world bet on the victory of one or the other candidate. The results are amusing: the odds of each of the competitors evolve according to campaign events. Joe Biden misses his debate? Hop, his curve collapses. Donald Trump miraculously survives assassination attempt? His rating soars and reaches its highest level of the entire presidential campaign. These sometimes surprising predictions must obviously be viewed with suspicion since they are only based on the personal feelings of the bettors. But we must recognize that they have taken on an insane scale. On the Polymarket platform, more than $1.5 billion in bets have been made. “At the end of August, this bet had the honor of being integrated into the Bloomberg financial terminal in the middle of surveys from prestigious institutes,” even underlines our columnist Robin Rivaton.
The platform recently hired a top advisor: Nate Silver. He, who is usually quick to defend his model, prefers to play it safe. In one of his last postshe gives a slight ascendancy to Harris, 55 to 45%. “That means Harris has an 11 in 20 chance of winning the electoral college and Trump has a 9 in 20 chance,” he explains. “It’s like flipping a coin, with the coin barely unbalanced in favor of Kamala Harris.” With such predictions, the model is unlikely to be wrong. But then why its own coast Polymarket is it so low?
.