Milei: 51.1%. Mass: 48.8%. It remains the average of the last fifteen polls recorded in Argentina before the second round of the closest and at the same time most atypical presidential elections that Argentina can remember. The ruling party’s successor candidate and the outsider on the right are virtually neck and neck if you leave the decimals aside and contextualize the two estimates that are close to above or below 50% with an average that eliminates the undecided and undecided from its calculation base . A blank vote was registered in the polls to bring his photo closer to the image of what we could see this Sunday.
The difference of 2.3 percentage points is within the margin of error of most polls published not only in Argentina but in virtually every country. It would be enough for a victory change with a half-value change of 1.2 in favor of one of the candidates, which would automatically be against the other. There are very few surveys that, on paper, manage to reduce the margin of error below “more or less” (addition or subtraction) of 1.2 points. Almost none of them make it believable outside of the role. And no average created with this commodity can logically significantly reduce this window of uncertainty.
In fact, a disaggregated analysis of the fifteen polls that average shows that the frequency of absolute differences between candidates above 2.3 is high. What actually happens is that nine of them awarded the win to Javier Milei, but six to Sergio Massa. The differences range from 0.8 to 9.9, but in both directions: The polling stations as a whole do not draw a similar race, close, but with a slight advantage for Milei as on average. No: The competition photos differ significantly from each other, which only increases the uncertainty of a forecast, not reduces it.
This is because when making a forecast using different sources of information, reaching consensus between them is completely different than such disagreements. Given this significant variance, the arithmetic mean cannot and must not be interpreted without it. This 51.1% from Milei rises to 54.9%, but also falls to 48.4%. Likewise, the average of 48.8% in Massa is 45.1%, but the upper limit is above 51% in more than one survey. These are the real areas of doubt before Sunday.
It is true that the poll numbers still fall more on one side (Milei) than on the other (Massa). But at the same time, the biggest error between the averages of the first round and the final result at that time was precisely the underestimation of the official candidate. In PASO the error was with Milei. But the methodological adjustments between then and now were more practical and credible because of the extended time (from August to October) and because the nature of the election was significantly different. So if probabilities had to be assigned between potential errors (in favor of Milei, in favor of Massa), the rightward bias might be slightly more credible.
In any case, these three sources of doubt (the narrow range of the average, its varying composition from survey to survey, and the first-round error) advise caution in interpreting the current average as a likely outcome. In fact, it’s more than a picture of how close this election is, it’s more a portrait of the uncertainty that surrounds it.
Average methodology. This average takes into account the most recent version of each poll released by polling places, with some fields conducted November 1-13 and viewed as of Thursday afternoon, November 16.
In order to make the polls comparable to each other and to bring the final number as close as possible to the picture perceived by the polls, the average is calculated, excluding the undecided and blank votes from the calculation basis. So, for example, if a poll shows 50% for Candidate A, 40% for Candidate B, and 10% undecided, Candidates A and B’s percentages will be recalculated based on the total number of decided votes, which would be the case in this case 50+40=90. Candidate A would have 50/90=55.55%. Candidate B 40/50=44.44%. All surveys captured in their original versions are listed below.
Subscribe to continue reading
Read without limits
_