For this calculation to happen, Cox shows that three conditions must be present: the voter cares about the result (they do not vote just to express an identity), they know who is viable and who is not, and this information is shared by everyone — which, in modern democracies, is the role of polls. When the three ingredients combine, the dispute converges on the two names with real chances, and the others shrink. Cox calls this "strategic defection."
The last three Quaest rounds in São Paulo, Minas Gerais, Rio de Janeiro, Pernambuco and Distrito Federal show this mechanism in operation. I added, in each state, the voting intentions of Augusto Cury (Avante), Ronaldo Caiado (PSD), Romeu Zema (Novo) and Renan Santos (Missão) — the four names competing for space outside the polarization — and compared them with the evolution of the advantage of whoever leads the first round in the last three rounds (8, 23 and 29 September). The pattern is consistent: the third way withers and the vote migrates to the candidate who was already ahead in that territory.
In four of the five states, the fall of the third way turns into a greater advantage for those who were already leading; in Minas, both are growing equally and President Lula's (PT) lead is 4 points.
In Pernambuco, the bloc of non-polarized candidates went from 10% on September 8 to 6% now. In the same period, Lula rose from 51% to 58%, while Flávio Bolsonaro (PL) fluctuated from 21% to 20%. The president's advantage, which was 30 points, was 38. The Pernambuco voter who tried Cury in September returned to his original vote.
In Rio de Janeiro, the movement is the mirror. The third way went from 12% to 8%; Flávio Bolsonaro went from 34% to 38% and Lula fluctuated from 32% to 31%. The senator's advantage, which was 2 points, reached 7. In São Paulo, the bloc went from 15% to 13%, Flávio rose five points (from 31% to 36%) and Lula stayed where he was, at 30%. The Bolsonarist lead went from 1 to 6 points.
In the Federal District the case is clearest. Caiado, which had 13% in the federal capital, today has 8%, and the third copy as a whole fell from 24% to 16%. Flávio went from 31% to 38% and opened 6 points over Lula, who rose less.
Minas Gerais is the exception that confirms Cox's logic. The third way deflated from 19% to 14% after Augusto Cury's peak, but Lula and Flávio grew the same four points, and the president's advantage, of 4 points, was only consolidated. Minas remains the most balanced state of the five — and, therefore, the one in which the useful vote has not yet decided the first round. Without being clear about who leads, a useful vote does not produce an advantage for either side.
What unites the five states is not the direction of movement, but its logic. The useful vote has no side: it strengthens those who are already leading where the voter perceives that the dispute is defined. In the Northeast, this expands Lula. In Rio, São Paulo and Brasília, Flávio Bolsonaro expands. The third way, which in September reached almost a quarter of the votes in the Federal District and a fifth in Minas, is being drained by both ends of polarization, exactly as Cox predicted almost thirty years ago.
For candidates outside the polarization, the lesson is harsh. Cox shows that strategic defection accelerates as the election approaches and polls make the hierarchy of candidates public consensus. With little to go until the first round, each new round of research that confirms the distance between the two leaders and the others works as a signal to voters: insisting on their favorite has a cost. And the Brazilian voter, as the numbers show, is doing the math.
