An election that changed shape overnight
On the night of 31 May, Colombia chose its two finalists and, in passing, contradicted nearly everyone who had forecast it. According to results released by the National Civil Registry, Abelardo de la Espriella took 43.74 percent — more than 10.3 million votes — and Iván Cepeda 40.90 percent, close to 9.7 million, a gap of just under three points. None of the thirteen candidates reached the 50 percent the Constitution requires to win in the first round, so the runoff was set for 21 June. Third and fourth went to Paloma Valencia and Sergio Fajardo, now out of the contest but not out of the equation, because their votes are the raw material the two campaigns are now fighting over.
The fact worth pinning down before going further is that the front-runner in the pre-election polls lost the first round. Until weeks before the vote, Cepeda led the surveys of five pollsters and his campaign saw a first-round win as within reach. He finished second. That single reversal forces a cautious reading of the runoff polls: the firms now projecting 21 June are, in large part, the same ones that got the order wrong on 31 May. The exception everyone cites is AtlasIntel, which is where to start.
The two readings that do not match
Here is the first tension in the data. There are two ways to measure a runoff, and they produce different numbers.
The first is to ask directly whom people would vote for on 21 June. AtlasIntel, the firm that came closest to the first-round results, measured for Semana a voting intention of 50.3 percent for De la Espriella against 42.6 percent for Cepeda: a gap of 7.7 points. In that same reading, blank votes stood at 3.7 percent, null or “will not vote” at 0.5 percent, and “don’t know” or no answer at 2.9 percent. It is a wide lead, and it comes from the pollster with the best recent record.
The second way is to model how the eliminated candidates’ votes split, starting from the actual first-round result. La Silla Vacía’s calculator, which distributes the other candidates’ voters, gives 50.5 percent to De la Espriella and 47.9 percent to Cepeda: a lead of 2.7 points. That reading describes a much tighter race than AtlasIntel’s.
The gap between 7.7 and 2.7 points is not a minor methodological footnote: it is the difference between a settled election and an open one. The source of the discrepancy is the assumption about transfer. A direct-intention poll captures the mood of the moment; a transfer model starts from the counted votes and applies distribution rules. When the two methods diverge this much, what they are saying is that the voters of the eliminated candidates have not fully decided, and that small variations in that split move the result.
Variable one: where the eliminated votes go
The first measurable factor is transfer. The first round left almost six million votes spread among Valencia, Fajardo and the rest. The question is where they move.
Paloma Valencia’s case is the clearest. The Centro Democrático candidate, third with about 6.9 percent, announced her support for De la Espriella the day after the vote, despite past quarrels. That endorsement steers a bloc of the right toward the front-runner. Citi Research, drawing on the first-round results, projected that De la Espriella would win in 17 of 18 scenarios analyzed, and that his largest reserve of new voters could come precisely from Valencia’s electorate.
The rest is murkier. According to La Silla Vacía’s reading, among the contested voters blank votes dominate at 43.6 percent; 24.2 percent would go with the Pacto Histórico and 18.9 percent with the harder right. That 43.6 percent now sheltering in the blank vote is the real battlefield: if it stays home or spoils, it helps whoever is ahead; if it breaks, it can narrow or widen the gap. La Silla Vacía’s transfer count ends at a 2.7-point lead precisely because that undecided bloc does not split cleanly.
There is a precedent both campaigns watch out of the corner of their eye. In 2022, the runoff preliminary count initially showed Rodolfo Hernández as the virtual winner, and the picture reversed as the official count advanced until Gustavo Petro prevailed with 50.51 percent against 47.22 percent, a gap of about 700,000 votes out of more than 22 million cast. That runoff also began looking like one thing and ended as another. The lesson of 2022 is not that the trailing candidate always comes back, but that transfer and turnout can move several points in three weeks.
Variable two: how many new people vote
The second factor is mobilization, and it is the one that introduces the most uncertainty. Citi described the runoff as “primarily a mobilization contest,” in which De la Espriella keeps the lead even if additional turnout favors Cepeda. The phrase carries a strong hypothesis: that the right-wing candidate has enough cushion to withstand a surge in progressive turnout.
The regional map helps locate where that mobilization would play out. According to AtlasIntel, De la Espriella would take the Caribbean coast with 68.6 percent, a region Cepeda had won in all eight of its departments in the first round, a historic stronghold of petrismo. That Caribbean swing, if confirmed, would be devastating for the Pacto Histórico, because there Cepeda gathered 2.2 million votes, close to 23 percent of his national total. In the opposite direction, Bogotá would stay with the progressive camp: in the capital Cepeda would hold 56.4 percent against 36.1 percent for De la Espriella. The right would also gain ground in Amazonía, Orinoquía and the center, where it would top 50 percent, the reverse of the first round.
The Bogotá detail deserves a note, because it tempers the Pacto’s optimism. Despite beating his rival by more than 160,000 votes in the capital in the first round, Cepeda did not match Petro’s 2022 performance nor reach his target of more than two million votes in Bogotá. In other words: even in its stronghold, the progressive camp mobilized less than four years ago. If that trend holds, the mobilization variable would work against Cepeda, not for him.
The cross-rejection, the figure that sets the tone
There is a third number that does not decide the election but explains its character. 56.6 percent of those surveyed reject Cepeda as a possible president, and 40.3 percent say the same of De la Espriella. Both are candidates with a high rejection ceiling, which sketches a contest between mutually exclusive projects. In an election like this, winning is less about thrilling your own electorate than about gathering the rejection of your opponent more effectively. Whoever can present themselves as the brake on the other has the edge.
To that logic is added the outgoing government’s approval, which weighs on the ruling-camp candidacy. In the AtlasIntel reading, 53.2 percent disapprove of Gustavo Petro’s record against 42.7 percent who approve. For Cepeda, who runs as the continuity of the Pacto Histórico project, that majority disapproval is a drag he must offset with his own turnout. For De la Espriella, it is ammunition: turning the runoff into a plebiscite on the outgoing government favors him.
The fine arithmetic of the eliminated
To grasp the transfer fight it helps to put exact figures on what fell away. The first-round breakdown left Paloma Valencia and Juan Daniel Oviedo with 6.92 percent, Sergio Fajardo and Edna Bonilla with 4.26 percent, Claudia López and Leonardo Huerta with 0.95 percent, Santiago Botero and Carlos Cuevas with 0.87 percent, and the blank vote at 1.72 percent. Added up, those blocs represent a little over thirteen points of the electorate that voted for someone no longer on the ballot, plus the blank vote. That is the caudal in dispute.
The Valencia-Oviedo ticket is the piece that already moved, and it moved to the right. But the other blocs are not automatic. Fajardo and Bonilla’s — the reformist center — is the most courted and the most elusive. Sectors close to Cepeda have sought to approach Sergio Fajardo, Juan Daniel Oviedo and Claudia López for the runoff, aware that neither De la Espriella nor Cepeda wins without the center’s votes. The obstacle is one of political memory: figures from that camp have recalled the moments when petrismo dismissed them as “lukewarm,” and neither Fajardo nor López seemed willing to align quickly with either finalist. Until that negotiation crystallizes, center voters may stay undecided for the full three weeks.
The detail to watch is that the center does not vote as a bloc nor obey its leaders. It is one thing for Fajardo or López to hint at a preference, and another for their electorate to follow it. In 2022, much of the center vote ended up fragmenting among abstention, the blank vote and the two extremes, with no leader managing to “deliver” it. The 2026 question is whether that pattern repeats or whether extreme polarization — two candidates with high rejection ceilings — pushes the center to choose the lesser evil with more discipline than four years ago.
The turnout record and what it means
There is a first-round figure that reorders the whole mobilization analysis: turnout was the highest in the country’s history. With 99.99 percent of tables reported, 23,978,053 Colombians voted, 57.88 percent of the 41,421,973 eligible, the largest turnout in a presidential election in all of Colombia’s history. De la Espriella also became the first Colombian to surpass ten million votes in a presidential first round.
That record changes the meaning of the mobilization variable. In an election with high abstention, the room to grow lies among those who did not vote, and the fight is about getting them out of the house. But if 58 percent already voted — a historic ceiling — the space to mobilize new people is narrower than usual. Colombia’s runoff tends to draw turnout similar to or slightly below the first round; in 2022 it was around 58 percent in both rounds. If in 2026 the first round already hit the historic ceiling, the runoff will be fought less on bringing in new voters and more on whom the already-mobilized are convinced by, and on which side demobilizes its base less.
It is worth placing the moment in institutional perspective. Since the 1991 Constitution introduced the runoff, only Álvaro Uribe has managed to avoid it, in 2002 and 2006; every election since has been decided in a second round. That 2026 goes to a runoff is no anomaly, it is the norm of the past two decades. What is unprecedented is the combination of extreme polarization, record turnout and a reversal of the front-runner in the first round. That mix is what makes any categorical projection risky.
The dispute over the count and institutional health
The close of the first round left an episode worth recording precisely, because it touches trust in the system. President Petro and then Cepeda himself questioned the preliminary count without presenting evidence and asked to wait for the ruling of the panels of judges that carry out the binding scrutiny. The Registry, for its part, dismantled the fraud allegations one by one and defended the regularity of the day.
The technical distinction matters. The preliminary count is an informational mechanism updated with data transmitted from the tables; the result with legal effect comes from the scrutiny the panels perform from the E-14 forms. The two readings usually agree on the trend, but only the second is binding. Questioning the preliminary count without evidence is not the same as proving fraud, just as defending the regularity does not foreclose any legitimate complaint by decree. The reportable fact is that there were allegations with no evidence supplied and a response from the electoral authority; the rest is political dispute.
The international component added noise to the close. Ecuador’s president, Daniel Noboa, and the president of the United States, Donald Trump, congratulated De la Espriella on the first-round result. De la Espriella, before a crowd in Barranquilla, said he would not allow “the people’s will to be stolen” and that he would defend democracy “by reason or by force.” That last phrase, by its tone, sets the confrontational register with which the right enters the runoff.
What to watch over the three weeks
From the pile of figures, four concrete indicators emerge to follow until 21 June, all measurable and all above the level of speeches.
The first is the consolidation or thaw of the blank vote among the eliminated candidates’ voters. As long as that 43.6 percent stays in limbo, the election remains more open than AtlasIntel suggests. If it starts to break, the direction of that move will be the best signal of where the result is heading.
The second is the Caribbean. It is the region where the two readings most contradict the first round, and where a real change in behavior would have the largest arithmetic effect. If local polls confirm the swing toward De la Espriella, the national lead becomes hard to reverse; if petrismo retains its stronghold, the election tightens.
The third is turnout in Bogotá and the big cities, the thermometer of progressive mobilization. Cepeda’s goal is not to win the capital — he already does — but to win it by the margin he needs to offset losses elsewhere. The number to watch is not who wins Bogotá, but by how much.
The fourth is the convergence or divergence among pollsters. If in the coming surveys AtlasIntel and the transfer models start to close in on each other, there will be a sign of consolidation. If they stay five points apart, the conclusion is that no one yet has a firm reading and that the result depends on what happens in the final week.
The balance of the numbers
None of the available figures closes the election, but all point in the same general direction: De la Espriella arrives ahead, with a lead that swings between the nearly eight points of direct intention and the under three of the transfer models. The Córdoba lawyer has in his favor Valencia’s backing, the Caribbean swing the polls sketch, the majority disapproval of the outgoing government and a lower rejection ceiling than his rival’s. Cepeda has in his favor that the real first-round gap was under three points, that the polls already got it wrong against him once — and could do so again, this time in reverse — and that the undecided blank vote is not yet distributed.
Colombia’s 21 June runoff will not be decided by who delivers the best speech in Barranquilla or Bogotá, but by two arithmetic operations: how many of Valencia’s and Fajardo’s votes end up in each ballot box, and how many people who did not vote on 31 May turn up three weeks later. Both are measurable, both are in dispute, and from both will come the name of the next president.