three weeks before the election, the polls more uncertain than ever

Data Regardless of opinion polls at the national level, the American presidential election, which is being held this year on November 5, will undoubtedly be played out in seven pivotal states where the gaps have narrowed in recent days between the Republican and the Democrat.

Three polls, three scenarios: October 8, NBC News places Kamala Harris tied with Donald Trump with 48% of voting intentions. The next day, Rasmussen Reports gave Donald Trump in the lead at 48%, ahead of the Democrat by two points (46%). Finally, on October 11, CBS News drew up an opposite plan, with Kamala Harris ahead of the Republican (51% against 48%). Welcome to the jungle of opinion surveys across the Atlantic, three weeks before the American presidential election!

Enough to lose the French spectator, accustomed moreover to a relatively simple, direct and single-member ballot in two rounds, when the election of the tenant of the White House, which takes place this year on November 5, is indirect, with a single turn… but above all, let us remember, is played out at the state level!

At the national level, Trump is struggling

For a French voter, the national poll is, in fact, a beacon on election night. This is much less the case for the American vote, due to the different institutes (and their possible political biases) but also to the political realities specific to each of the 50 states (and to the District of Columbia, also represented in the electoral college). , more complex to perceive with surveys limited to a few thousand respondents.

This is why we decided, at “New Obs”, to try to attenuate these oscillations by using the figures from two American platforms which specialize in the aggregation of polls: RealClearPolling, dependent on the political information site RealClearPolitics , rather classified on the Republican side, particularly since 2017, and FiveThirtyEight (538, being the number of electors who form the electoral college during the American presidential election), attached to ABC News, probably leaning slightly to the left. Their results differ somewhat. We can obviously see an influence of their political orientation, but the two aggregators also have different calculation methods and do not necessarily use the same polling institutes. FiveThirtyEight, in particular, will weight the different surveys according to those carried out in the past by the same institute and the results ultimately obtained. It also takes into account the political orientation of pollsters, etc.

Her two platforms currently give Kamala Harris in the lead, with 48.9% against 47.2% for the first as of October 13 (i.e. 1.7 points ahead) and with 48.5% against 46% as of October 12 for the second (2.5 points ahead).

A comfortable lead for the Democrat which dates back to around mid-August and which does not seem to have declined since.

If we can reasonably deduce that Kamala Harris should win the popular vote, the “Clinton” jurisprudence can only lead us to be cautious. In 2016, the former Democratic candidate for the White House had indeed won the most votes… but had nevertheless lost to a certain… Donald Trump (a scenario which, in total, happened five times in American history).

Because the American election is not played out at the national level, but at the level of each state, whose victory allows the candidate to win a certain number of electors (a number which varies depending on the state, from 54 for California to 3 for Wyoming), which then formally designate the American president.

The election will be played in seven “swing States”

Let’s then look again at the polls, but at the level of each state… Overall, poll aggregation platforms, like news sites, have been depicting a similar situation for months: Donald Trump would win in 24 states , Kamala Harris in 19 and Washington District. The only notable difference: Minnesota is considered by RCP as a pivotal state, without us really understanding why because the aggregator gives the Democrat nearly 5 points ahead of the Republican in this Midwestern state including her running mate, Tim Walz, is the governor…

Seven states are therefore ultimately presented as undecided. These are the famous swing Stateswhich can tilt from one side to the other. They often vary between two elections, or even sometimes during the campaign. This time, they haven’t changed in months…

Based on the previous maps, here is the number of electors that we can, fictitiously, “attribute” to each of the candidates, according to the two aggregators.

Whatever the aggregator, the balance is almost perfect: to the benefit of Donald Trump for RCP (219 voters, against 215 for Kamala Harris) or, on the contrary, with a slight lead for the Democrat for ABC News (226, against 219 voters for the Republican). Remaining 104 or 93 electors in the swing States…

The gaps are narrowing

So what is the current situation in these famous pivotal states? Let’s get rid of the Minnesota case straight away. For the other seven states, the two aggregators now display minimal differences, never exceeding more than one point (with the exception of Arizona, in favor of Trump, at 538). We are more than ever within the margin of error of the polls. However, the two platforms agree to award Arizona, Georgia and North Carolina to the Republican billionaire, while they only agree on Wisconsin in favor of the Democrat…

For the three other states, the results differ: RCP foresees a turnaround in Michigan and Nevada in favor of Donald Trump and a narrow victory for the Republican in Pennsylvania, while 538 instead envisages a victory of a narrow head. Kamala Harris in these three states…

Obviously, the final scenario is not the same: with the RCP figures, Donald Trump would win by a large margin, with 302 voters compared to only 236 for Kamala Harris. History would take a different turn if the results stick to the estimates of 538, since Joe Biden’s current vice-president would be the first black woman elected president of the United States, with 276 voters against 262 for the Republican.