Who will win Colombia's 2026 presidential election (first round 31 May, runoff 21 June)?

TL;DR

Abelardo de la Espriella is the clear favorite: he won the first round, faces only Iván Cepeda in the runoff, and the first post-round polls plus transfer math put him about seven points ahead.

Context

Colombia is now in a two-candidate runoff. The first-round official scrutiny reported by the CNE on June 4 put Abelardo de la Espriella first with 10,366,143 votes, or 43.78%, and Iván Cepeda second with 9,703,921 votes, or 40.98%; Paloma Valencia finished third with 1,640,989 votes and Sergio Fajardo fourth with 1,009,346 votes (Infobae, June 4). The Registraduría published the June 21 runoff ballot on June 5 with only two formulas: Iván Cepeda/Aída Quilcué and Abelardo de la Espriella/José Manuel Restrepo (Registraduría, June 5).

That makes Valencia, Fajardo, and the rest live only through rare replacement, disqualification, death, or deadline paths. Blank vote is not a route to a new election in a presidential runoff; the Registraduría says it has symbolic but not legal effect in the second round, and the candidate with more votes wins (Registraduría, June 15 2022).

Evidence

The historical backbone favors the first-round leader, but not enough to close the case. Since the 1991 constitution created the two-round presidential system, Colombia has had six presidential runoffs before 2026; the first-round leader won four and lost two, with reversals in 1998 and 2014 (Radio Nacional, June 19 2022). The full runoff history is:

De la Espriella’s 2.80-point first-round lead is not huge, but the pool of transferable votes is unfavorable to Cepeda. Using the CNE-reported counts, the eliminated non-blank candidate pool is 3,198,688 votes, and Cepeda starts 662,222 votes behind; if turnout were unchanged and all eliminated-candidate voters chose a finalist, Cepeda would need 60.4% of that pool just to tie (Infobae, June 4). That is a high bar because the largest eliminated bloc is Valencia’s right-wing electorate.

The first post-round polls point the same way. AtlasIntel/Semana, fielded June 1–2 with 2,030 respondents, Atlas Random Digital Recruitment, and a stated ±2-point margin of error, put De la Espriella at 50.3%, Cepeda at 42.6%, blank at 3.7%, undecided at 2.9%, and null or not voting at 0.5% (Semana, June 2). CB Global Data, fielded June 1–4 with 1,475 adults, online CAWI recruitment, and a stated ±2.6-point margin of error, put De la Espriella at 49.6%, Cepeda at 43.3%, blank/null at 3.6%, and undecided at 3.5% (El País Cali, June 5). A sample-size-weighted average of those two polls gives De la Espriella a 7.1-point lead over Cepeda.

The transfer cross-tabs make the polling lead look grounded rather than random. Atlas reported that 76.4% of Valencia voters would choose De la Espriella and 4.4% would choose Cepeda; among Fajardo voters, 18.9% would choose De la Espriella, 24.2% Cepeda, and 43.6% blank; among voters for other eliminated candidates, 53.2% would choose De la Espriella and 32.8% Cepeda (Semana, June 2). Applying those rates to the official first-round counts gives about 12.10 million votes for De la Espriella and 10.20 million for Cepeda before any new turnout effect, or roughly 54.3% for De la Espriella among the two candidate votes (Infobae, June 4; Semana, June 2).

Endorsements also favor De la Espriella on net. Valencia endorsed him on election night after finishing third, while Fajardo did not endorse either finalist and instead asked the campaigns to reject polarization and meet institutional conditions (Infobae, May 31; El País, June 3). I model this as a normal runoff in which De la Espriella is favored by the current polls and transfer math, but with wide error bands: direct polling at about 84%, transfer arithmetic at about 84%, and history/fundamentals at about 75%. Weighting those 45%, 35%, and 20%, then reserving about 0.6% for abnormal resolution paths, gives De la Espriella about 82% and Cepeda about 17%.

What's non-obvious

The race is not as close as the raw 2.8-point first-round gap makes it look. Cepeda’s main path is not simply to win the center; it is to mobilize new voters while limiting anti-Petro consolidation. El País reported a data analysis showing more than 1 million abstainers in Cepeda-favorable territories versus about 600,000 in De la Espriella-favorable territories, but the same article described weak Cepeda performance in Bogotá’s middle-class zones and said De la Espriella won eight localities that voted mainly for Petro in 2022 (El País, June 7).

Cepeda is trying the right repair strategy: he is distancing himself from Petro’s constituent-assembly push and aiming at youth, center voters, and abstainers, while De la Espriella is using his vice-presidential pick José Manuel Restrepo to sound more economically credible (El País, June 7). That keeps Cepeda live. But with only two weeks left, he needs a real shift, not just a better message.

Limitations

The biggest weakness is polling quality. I found only two public post-first-round runoff polls with enough detail to use, and both rely on online recruitment; reported sampling error is much smaller than the real forecast error. Pre-first-round polling also missed the scale of De la Espriella’s surge, with AS/COA summarizing that he outperformed polls and finished first with 43.7% while Cepeda had 40.9% (AS/COA, June 4). A credible face-to-face post-round poll showing the margin below three points would move this forecast down.

The administrative tail is small but not zero. The Registraduría said on June 2 that judicial scrutiny was 99.98% complete, with only 33 of more than 122,000 tables pending, and that the pre-count matched the scrutiny at 99.94% (Registraduría, June 2). Article 190 still creates rare replacement or postponement paths if a runoff finalist dies or becomes permanently incapacitated before the second round (Constitute Project, Article 190). After forming this estimate, I checked the market anchor mentioned in the prompt; Polymarket was roughly 81% for De la Espriella after the first round, close to this forecast, so I made no adjustment (Polymarket, June 7).