Public and private AtlasIntel polls accurately captured the dynamics of voter mobilization and came closer to the final vote count than any competitor.

• AtlasIntel public releases both before and after the start of the poll blackout period indicated that slightly more than 58% of Chilean voters would vote to reject the new constitution; with 61,8% of votes cast against the new constitutional text, AtlasIntel came closest to the final outcome regardless of whether one considers polls released before or during the blackout period.
• Unlike Cadem and other major Chilean polling houses, AtlasIntel did not indicate a trend reversal in the final days of the campaign for the approval of the constitutional text, let alone any chances for the “approve” option to prevail; although Atlas Tracking PRO indicated a steady rise in the share of Chileans intending to vote “approve” during the six weeks before the day of the vote, this rise failed to make a significant dent in the huge advantage of the “reject” vote (Atlas Tracking PRO time series presented below).

The New Constitution failed to be adopted because it lost adherents even among those segments most likely to support it, in addition to a much more uniform pattern of turnout mobilization in comparison to the last presidential election

• AtlasIntel crosstabs indicate that a sizeable share of Boric, Provoste, and Ominami voters rejected the constitutional text; as such, the new text was not even a consensus among the progressive voters that were most inclined to support it, given that they had voted massively in support of the Constitutional Assembly in 2020.
• Massive demonstrations in favor of the new constitution in the last stage of campaigning may have stimulated a counterwave: as voters fearful that the new constitutional text had chances of getting approved, voter turnout approached record levels.
• Unlike in the runoff of the 2021 presidential election, mobilization of voters was more balanced across demographics, with male and older voters turning out in much greater numbers and thus contributing to a slightly better result for the “reject” vote than anticipated by AtlasIntel final estimates.