Manila Voting Weather

Research paper

Manila Voting Weather of 2025: Why Pre-Election Surveys Failed to Predict the Midterm Outcome in the City of Manila

Allen Bornasal, Nathaniel Cachuela Jr., Almhea Irra Celeste, Jericho Jonas Franco, Christian Sabarre, Arianne Renae Santos, and Erwin Jasper Shi

DATAVIS BTIS1, De La Salle – College of Saint Benilde · 3rd Term, AY 2024-2025 · Prof. Bob Mathew Sunga

Abstract. The May 2025 midterm elections in the City of Manila produced two results that pre-election surveys did not anticipate: a landslide mayoral victory for Isko Moreno Domagoso, who won 822 of the city's 895 barangays with 59% of the vote, and a first-place senate finish for opposition candidate Bam Aquino, who was marked on the ballots of roughly 55% of Manila voters. Using a barangay-level dataset of 8.5 million senate ballot marks and 897,320 mayoral votes, this paper measures both results, compares them against campus mock polls and commercial survey findings, and examines why the polling instruments missed. Three mechanisms account for most of the gap: 62% of voters decided within a week of the election, after most instruments had stopped measuring; the mock polls that circulated during the campaign were convenience samples of student populations rather than probability samples of the electorate; and the coordinated Isko and Atienza tandem, together with Aquino's uniformly distributed support, produced momentum effects that static, name-recognition polling does not capture.

1 · Introduction

Pre-election surveys are supposed to reduce uncertainty. In Manila in 2025 they added to it. National polling placed Bam Aquino outside the top three for much of the campaign; he finished first in the capital. Commentary treated the mayoral race between a former mayor and a sitting mayor as competitive; it ended 59% to 21%, with the incumbent losing 822 of 895 barangays. The gap between what the instruments said and what the ballots said is the subject of this paper.

Three questions structure the analysis:

  1. How was Isko Moreno's victory distributed across the city's barangays, and what does that distribution say about its causes?
  2. What made Bam Aquino's first-place finish possible in a city where his national polling did not predict it?
  3. Which specific properties of the 2025 polling instruments caused them to miss both results?

The unit of analysis is the barangay, the smallest administrative unit in the Philippines. Manila's 895 barangay-level returns are granular enough to distinguish a candidate whose support is concentrated in strongholds from one whose support is spread thin and wide. That distinction turns out to carry most of the answer to question two.

2 · Background

2.1 The races

The 2025 midterms put three contests on every Manila ballot: twelve senate seats elected nationwide (each voter may mark up to twelve names), the city mayoralty, and the vice mayoralty. Isko Moreno Domagoso, mayor from 2019 to 2022, sought to return against incumbent Honey Lacuna of the Asenso Manileño coalition. Moreno ran under Aksyon Demokratiko in tandem with vice-mayoral candidate Chi Atienza. In the senate race, Bam Aquino ran as an opposition candidate under KNP after topping youth-preference surveys but polling outside the leaders in general-population instruments (The Global Filipino Magazine, 2024).

2.2 Why Manila might behave differently

Urban centers, and Manila in particular, have historically leaned toward opposition candidates in politically charged cycles; the Manila Times characterized the 2025 opposition surge as part of a broader protest vote against the sitting administration (Malindog-Uy, 2025). At the same time, national politics remained dominated by dynastic families: after the 2025 midterms, investigative counts put 71 of 82 provincial governorships in the hands of political clans (Latoza, 2025). Manila's choice of a non-dynastic mayoral candidate and an opposition senate leader is notable against that backdrop, and part of what the surveys had no prior for.

2.3 The incumbent's position

The Lacuna administration entered the campaign carrying a publicized 10 billion peso debt controversy, including 3.8 billion pesos in cash advances by special disbursement officers over ten months and 2.1 billion pesos advanced by 21 officials during the January to May 2025 election period (Quodala, 2025). Moreno made this record the center of a performance-based campaign, contrasting it with the infrastructure and rehabilitation program of his own 2019 to 2022 term (Taculao, 2022; Escosio, 2025).

3 · Data and methods

3.1 Dataset

The tally data is a publicly shared, nationwide scrape of the 2025 Philippine midterm election results (2025 Philippine Midterm Elections Data, 2025). The team filtered the national file to the City of Manila, yielding 895 barangay rows covering three races: approximately 8.5 million senate ballot marks, 897,320 mayoral votes, and 874,410 vice-mayoral votes. Cleaning and modeling were done in Excel and Power BI; the interactive figures on this site are re-implementations of the same model in hand-coded SVG.

3.2 Measures

  • Vote share is a candidate's votes over all votes cast in that race within the dataset. For the senate race, two denominators are reported side by side: share of all ballot marks (8.5M), whose theoretical per-candidate ceiling on a twelve-pick ballot is one twelfth, or 8.33%; and share of voters, approximated by the mayoral turnout of 897,320. Both are reported because each answers a different question.
  • Barangay win rate is the fraction of the 895 barangays in which a candidate received the most votes in their race.
  • Performance tiers classify each barangay by Moreno's local vote share into four mutually exclusive bands: Stronghold (60% and above), Competitive (45 to 59%), Swing (30 to 44%), and Challenging (15 to 29%).

3.3 Survey comparison

Campaign-period mock polls were collected from three university publications: the Artlets Student Council poll at UST (n = 182; Querol, 2025), Pulsong Adamsonian at Adamson University (n = 538; Pulsong Adamsonian, 2025), and the Benildean Halalan 2025 mock results (n = 512; The Benildean, n.d.). Their leading candidates were checked against Manila's actual senate top 15. Post-election diagnostics come from Pulse Asia's review of decision timing (Lalu, 2025).

4 · Results

4.1 The mayoral race

Moreno received 529,510 votes, 59% of the 897,320 cast. Lacuna received 190,320 (21.2%), a margin of 339,190. Sam Verzosa's 164,100 (18.3%) made the race effectively three-way; the remaining nine candidates shared under 2% between them. Moreno won 822 of 895 barangays, a 91.8% win rate.

The tier distribution locates the landslide everywhere at once: 360 barangays were Strongholds at 60% or more of the local vote, 411 were Competitive wins at 45 to 59%, 117 were Swing at 30 to 44%, and only 7 of 895 were Challenging at under 30%. Combining the two upper bands, Moreno took at least 45% of the vote in 771 barangays, 86% of the city.

Figure 1. Barangays by Moreno vote-share band, mutually exclusive, n = 895. An interactive version with a table view is in the dashboard.

4.2 The vice-mayoral race

Chi Atienza received approximately 582,500 of 874,410 votes, 66.6%, against 249,400 (28.5%) for incumbent vice mayor Yul Servo Nieto. Atienza's share exceeds Moreno's own 59%: the tandem's junior partner outperformed its principal, which is the clearest single measurement of ticket voting in the dataset.

4.3 The senate race

Aquino led Manila with 498,150 ballot marks. Expressed against all 8.5 million marks this is 5.86%, which understates the result because a twelve-pick ballot caps any candidate at 8.33% of marks; Aquino took 70% of the attainable maximum. Expressed against voters, roughly 897,000, his 498,150 marks mean approximately 55% of Manila voters included him among their twelve choices. Bong Go followed at approximately 426,600 marks, then Kiko Pangilinan at 365,400. Two of the top three, Aquino and Pangilinan, were opposition candidates.

Aquino won 775 of 895 barangays, an 86.6% win rate, with a best single barangay of only 793 votes. The distribution has no strongholds: his barangay-level share of marks sits between 4 and 6% nearly everywhere. The upset was built out of thin, uniform support across the whole city rather than deep support anywhere.

4.4 Surveys against outcomes

The campus mock polls pointed at a different election. At Adamson, Aquino polled 93.49% and Pangilinan 92.94%; at UST Artlets, the leading names included Danilo Ramos (63.7%), Arlene Brosas (62.1%), and Teddy Casiño (59.3%). Of nine named mock-poll leaders, only Aquino and Pangilinan appear in Manila's actual top 15. The instruments were directionally right about Aquino and Pangilinan and wrong about nearly everything else, which is the signature of a sample that measures a real but unrepresentative population.

Figure 2. Timing of senate vote decisions, from Pulse Asia's post-election review (Lalu, 2025). The three right-hand bands, 62% of voters, decided within a week of the election.

Pulse Asia's own post-election diagnosis quantifies the second failure mode: 26% of voters settled their senate slate a week before the election, 25% the day before, and 11% on election day itself. Only 38% of final decisions existed early enough for a campaign-period survey to observe them.

5 · Discussion

5.1 Why the landslide

Three factors, each visible in the data or the record:

  • A performance contrast voters could verify. Moreno campaigned on the visible record of his 2019 to 2022 infrastructure program (Taculao, 2022) against an incumbent carrying a 10 billion peso debt controversy (Quodala, 2025). The uniformity of his tier distribution, at least 45% in 86% of barangays, fits a citywide judgment rather than a targeted one.
  • A coordinated tandem. Moreno and Atienza ran as a unit under Aksyon Demokratiko and were endorsed as a pair, including by the bloc-voting Iglesia ni Cristo (Nicol, 2025). Atienza's 66.6% exceeding Moreno's 59% measures how completely the ticket, not the individual, was the object of the vote.
  • A biography that carried. Moreno's poverty-to-office narrative gave the performance case an emotional anchor (Escosio, 2025). This is the component polls can in principle measure but rarely weight.

5.2 Why the upset

  • Youth preference realized as turnout. Aquino topped preference surveys among voters aged 18 to 25 well before election day (The Global Filipino Magazine, 2024); the campus mock polls, whatever their sampling defects, recorded the same signal at extreme strength.
  • A protest-vote city. Manila's historical opposition lean in charged cycles (Malindog-Uy, 2025) supplied a receptive electorate that national instruments, calibrated on national sentiment, discounted.
  • Barangay-level platform, barangay-level result. Aquino campaigned on benefits for barangay tanods and health workers: salaries, PhilHealth coverage, training, and legal support (GMA Integrated News, 2025). A platform aimed at every barangay's frontliners is a mechanism for exactly the thin, universal support the returns show.

5.3 Why the instruments missed

  1. Timing. With 62% of decisions made in the final week, any instrument that stopped interviewing earlier was measuring at most 38% of the outcome (Lalu, 2025).
  2. Sampling. The circulating mock polls were convenience samples of single campuses. They measured a real constituency, which is why they called Aquino and Pangilinan, and an unrepresentative one, which is why two thirds of their favourites missed the city's top 15 entirely.
  3. Coordination effects. Static polling reads candidates independently. The tandem's joint endorsement structure and Aquino's frontliner platform both generated late-compounding momentum that an independent-candidates model cannot represent.
  4. Spending is a bad proxy. The 2025 cycle's top spender, Camille Villar at 179.6 million pesos, did not convert expenditure into a proportional result (Fonbuena, 2025), removing the one covariate lazy forecasts lean on.

5.4 Looking to 2028

Manila's 2025 result is an early observation for the 2028 presidential cycle: the capital rewarded performance records and opposition positioning while the Marcos and Duterte blocs traded national ground (Robles, 2025; De Guzman, 2024). Commentators have raised both a Sara Duterte candidacy and the possibility of a dark horse from outside either dynasty (Butuyan, 2025). If the 2025 pattern holds, instruments that cannot see late deciders and coordination effects will misread 2028 the same way.

6 · Limitations

  • One missing barangay. The dataset carries 895 barangay rows; Manila is generally counted as 896. Every rate in this paper is therefore a rate over the dataset, not over the city. No conclusion changes at n = 896.
  • The senate total is rounded at source. The report states 8.50M marks; the share-of-marks figures inherit that rounding.
  • Voter count is approximated by mayoral turnout. The share-of-voters denominator assumes senate participation approximately equals mayoral participation. Undervoting between races (874K vice-mayoral against 897K mayoral votes) suggests the true senate-voter count differs by a small margin.
  • Ecological inference. All findings are barangay-level patterns. They constrain but do not identify individual voting behavior; statements about youth support rely on external surveys, not the tally.
  • Bloc labels are judgments. The opposition/administration/Duterte-aligned classification of senate candidates follows common press usage in the cycle and is not an official designation.

7 · Conclusion and recommendations

Manila's 2025 election was not unpredictable; it was unmeasured. The returns show a city-wide, verifiable judgment: a performance candidate swept 92% of barangays, his running mate outran him, and an opposition senator was marked by more than half the city's voters without a single stronghold. Each of those patterns has a mechanism visible in the record. What failed was instrumentation pointed at the wrong week and, in the case of the mock polls, the wrong population.

For future cycles, three concrete changes follow from the data:

  1. Measure the final week. With 62% of decisions made inside seven days, tracking-style instruments with election-week waves are not an upgrade but a minimum.
  2. Track local issues, not just names. The debt controversy and the tanod-benefits platform moved this election; neither appears in a name-recall questionnaire.
  3. Model coordination. Tandems, endorsements, and bloc votes make candidate outcomes dependent. Forecasts should carry explicit terms for ticket effects rather than treating each name as an independent draw.

References

APA 7th edition. Links are to the versions consulted during the project.