DPWH flood-control records

DPWH flood-control contracts and the firms that won them

A graph of every DPWH flood-control contract on the public record, the firms that won them, and the license revocations, blacklists, and court filings now attached to some. Each claim links to its source. Not verdicts, but statistical indicators drawn from the records.

Read the record Explore the graph

₱1.586T
in flood-control contracts: 2016 to 2026, recorded; 33,866
flood-control projects: DPWH category; 6,558
distinct firms: resolved from 248,220 contracts; 20
concentrated offices: one or two firms hold most of the budget

Data from public records (DPWH Transparency Portal via BetterGov.PH, CC0). This tool computes statistical indicators only. Charges are allegations under the presumption of innocence, and each links to its source.

Reading the record, in sequence

What the records show

01The scale
One category, ₱1.586 trillion
Between 2016 and 2026, DPWH recorded 33,866 flood-control and drainage contracts. By value it is the second-largest infrastructure category in the whole dataset, behind only roads.
₱1.586Trecorded flood-control value
02The surge
Spending tripled in six years
Flood-control allocations climbed from about ₱124 billion in infrastructure year 2018 to ₱368 billion in 2024, the peak year, before easing in 2025. Most of the contested projects sit inside that run-up.
₱368Bflood-control value, 2024 (the peak)
03Concentration
Twenty district offices with high concentration
In 20 district engineering offices, one or two firms hold most of the flood-control budget: a Herfindahl-Hirschman Index above 2,500, the US Justice Department line for a highly concentrated market. News reporting and the administration's own review found about 15 firms took roughly ₱100 billion, near 20% of the program.
20highly concentrated district offices
Flood control scandal (Wikipedia)
04The epicenter
Bulacan, where COA flagged unbuilt projects
Flood-control contracts in Bulacan's district offices total about ₱102.9 billion. This is where the Commission on Audit found projects paid for but unbuilt, and where the top awardees include Wawao Builders and Topnotch Catalyst Builders.
₱102.9Bflood-control value in Bulacan offices
COA fraud audits (Rappler) Open in the explorer
05The reach
A few firms across dozens of offices
A handful of contractors do not just win in one place. St. Gerrard Construction appears as a flood-control awardee across 96 district offices; St. Timothy across 82; Legacy across 87. In the graph they sit between clusters, with the widest reach across offices.
96district offices touched by one firm
Open in the explorer
06One family, nine firms
The Discaya companies
On September 1, 2025, Sarah Discaya told a Senate hearing she owned nine construction firms. Two days later the Philippine Contractors Accreditation Board revoked all nine licenses (Board Resolution 075) for collusion and bid-rigging. Several appear here as top flood-control awardees, and the DPWH dataset already marks their licenses as revoked.
9firms, licenses revoked by PCAB
9 Discaya firms stripped of license (Inquirer) Open in the explorer
07The first charges
Sunwest and Zaldy Co
In November 2025 the Ombudsman filed the first criminal charges in the flood-control case before the Sandiganbayan: graft and malversation against company directors and resigned representative Zaldy Co, over a ₱289.5-million flood-control project in Oriental Mindoro. Charges are allegations; the case is pending.
₱32.83BSunwest recorded flood-control value
Sandiganbayan cases vs Zaldy Co (GMA) Open in the explorer
08Flags and freezes
Audits, blacklists, frozen assets
The Commission on Audit flagged a Topnotch riverbank project in Bulacan. Separately, the anti-money-laundering council secured court orders freezing assets tied to the Discaya family. DPWH ordered the perpetual disqualification of Wawao Builders and SYMS. Even so, the eight firms carrying a revoked-license tag in the data won about ₱74.5 billion in flood control alone.
₱74.5Bflood-control won by revoked-license firms
DPWH bans Wawao, SYMS (Manila Times) Open in the explorer
09What this is
Not verdicts, but indicators
Everything here is a public record: a contract, a license status, a court filing, an audit report. The graph computes statistical indicators, who co-awards with whom, who won the most, where the money concentrates. It does not decide guilt. Charges are allegations under the presumption of innocence, and every claim links to its source.

Search-first

Explore the graph

Search 6,558 firms and 220 district offices by name, by an owner or officer named in the records, or by a former company name. Select a node to see its record, its recorded joint ventures, the official actions on file, and the sources.

Try: Topnotch, Sunwest, Legacy, or a district office.

Jump to:

112 nodes · 762 links

Highest recorded flood-control value

Sum of contract budgets where the firm is sole or joint awardee. Select a firm to see its record.

Network analysis, computed from the record

How the network formed

Year by year, 2016 to 2025: the money grew, the named firms' share of it grew, the market closed to newcomers, and the joint-venture web consolidated. Every figure below is computed from the contract record. Statistical indicators derived from public data. Patterns may have legitimate explanations.

0 firms & offices0 linkslargest group 0

Replay of the scandal-core network forming, 2016 to 2025: firms (circles) and district offices (squares) appear when they first connect, award links in grey, recorded joint ventures in ochre. Link width scales with the number of flood-control contracts; node size with recorded contract value; positions are fixed so the growth reads as densification. Node colour marks the record on a firm. Descriptive, from public data.

Flood-control value per year

From ₱52.60B (2016) to a ₱368.08B peak (2024).

The 16 named firms' share

Share of each year's flood-control value won by the firms later named in the inquiry: 5.3% to a 14.4% peak (2024).

Newcomers' share of the money

Value share of firms winning flood control for the first time. The market closed as spending tripled. (2016 is the observation start and is excluded.)

The joint-venture web

Firms in the largest connected group of the cumulative JV network: 26 (2016) to 1,392 (2025), one web.

Temporal knowledge graph, validated

Beyond the static graph: what the record shows once every award and joint venture carries its year, and every claim is tested against a null model. Statistical indicators derived from public data. Patterns may have legitimate explanations.

Predicting next year's partnerships

AUC 0.702

Train on the awards up to each year, then predict which firms form a new joint venture the year after, from their prior shared-office structure alone. Skill 0.702 on average, strongest in 2024 (AUC 0.761). It beats a degree-preserving null in every split (p < 0.01), so the signal is the structure, not just which firms are big.

Blue is real skill, dashed grey is the degree-preserving null. Predictions are statistical similarity, not evidence of a relationship.

The clusters consolidated

8 → 5

Co-location communities each year: they fell from 8 in 2018 to 5 by 2025, while the largest grew from 59 to 158 firms. The partition settled (year-over-year stability 0.946), so the structure is not churn: the same clusters persisted and merged.

Largest co-location community, firms, 2018 to 2025.

Sequence, not coincidence

For firm pairs holding a joint venture, the JV forms before their shared awards concentrate 263 times, versus 66 the other way, about 4.0x. The partnership tends to predate the money. Order only, never intent.

263 JV-firstvs66 awards-first

When the structure shifted

Pettitt change-point tests put the break around 2020 across several series. The named firms' share steps from 6.95% to 12.22% at 2020. Over a ten-year window (2016 to 2025) the significance is marginal (p ≈ 0.07), so read it as a candidate turning point, not proof.

The money-and-power schema

A heterogeneous temporal graph (person, firm, office, institution nodes; typed, dated edges) built from the sourced overlay: 44 edges across 8 people and 16 firms. Bulk SALN wealth and SOCE campaign finance are not machine-readable yet; this is the interface, filled with the real records that exist, ready to scale.

Patterns in the record

Structural patterns that the procurement-integrity literature (e.g. the OECD guidelines for fighting bid rigging in public procurement) associates with coordinated bidding. Each is a descriptive statistic with the threshold stated. None is a finding of wrongdoing; legitimate explanations include regional specialization, geography, and licence class.

Near-identical bidding footprints

64 found

Two firms with no recorded joint venture whose flood-control district-office footprints are near-identical (Jaccard >= 0.6, >= 4 shared offices; firms with >= 10 flood-control contracts and >= P500M).

G.C.I. CONSTRUCTION AND DEV'T. CORP + BICOL GOLDROCK CONSTRUCTION CORPORATION8 offices · J 0.667
F. GURREA CONSTRUCTION, INCORPORATED + IBC INTERNATIONAL BUILDERS CORPORATION10 offices · J 0.667
EIGHT J'S CONSTRUCTION SERVICES + JELSIE CONSTRUCTION & SUPPLY5 offices · J 0.625
MAKAPA CORPORATION + RAMARPLUS INCORPORATED5 offices · J 0.714
M3 KONSTRACT CORPORATION + 729 CONSTRUCTION, INC6 offices · J 0.6

Joint-venture rings

14 found

Groups of 3+ firms connected through recorded joint ventures on flood-control contracts (connected components of the co-award graph).

ROCK-WBL CONSTRUCTION CORP. (FOR: ROCK-WBL CONSTRUCTION TRADING & SERVICES), SILVERWOLVES CONSTRUCTION CORPORATION, UNIMAX STEEL STRUCTURE AND CONSTRUCTION CORP +25 firms · ₱10.69B
WEARR ENTERPRISES & DEVELOPMENT CORP, 6RM CONSTRUCTION AND DEVT. CORP, RCDS CONSTRUCTION & SUPPLY, INC.(FOR RCDS CONST +36 firms · ₱9.08B
A.R.L. CONSTRUCTION, TBB CONSTRUCTION CORP, AR LUSTRE CONSTRUCTION CORPORATION3 firms · ₱4.45B
KVC CONSTRUCTION, BAKUN CONSTRUCTION, GACOSCOS CONSTRUCTION +47 firms · ₱3.84B

Alternating top awardees

0 found

Highly concentrated offices (HHI > 2500) where the yearly top awardee alternates between exactly two firms across 4+ active years (>= 3 flood-control contracts per counted year, >= 2 switches).

None met the threshold. Reported as tested: concentration here shows up as one dominant firm per office, not two firms trading the top spot.

New firms, immediate large awards

16 found

Firms whose first flood-control award came in 2020 or later and that won >= P1B within their first two years in the category.

SILVERWOLVES CONSTRUCTION CORPORATION2023 · ₱6.53B in 2 yrs
ONE FRAME CONSTRUCTION INC2024 · ₱3.80B in 2 yrs
ZLREJ TRADING AND CONSTRUCTION CORPORATION2022 · ₱2.32B in 2 yrs
ELITE GENERAL CONTRACTOR AND DEVELOPMENT CORP2022 · ₱1.74B in 2 yrs
JJPR CONSTRUCTION SERVICES2023 · ₱1.52B in 2 yrs

Where the flood-control money went, by region

Region is the DPWH record's location.region. Canonical PSGC province and municipality codes are not joined: the PSA publishes PSGC only as manual downloads behind Cloudflare, so there is no automated join. A point-level map is a natural next step; the record carries latitude and longitude on most contracts.

Region III

₱267.14B · 16.8%

National Capital Region

₱157.85B · 10%

Region V

₱154.28B · 9.7%

Region IV-A

₱132.50B · 8.4%

Region I

₱108.94B · 6.9%

Region IV-B

₱97.92B · 6.2%

Region II

₱85.43B · 5.4%

Region VII

₱76.38B · 4.8%

Region VIII

₱76.19B · 4.8%

Region XI

₱64.32B · 4.1%

Region X

₱64.12B · 4%

Region XIII

₱51.98B · 3.3%

Cordillera Administrative Region

₱48.39B · 3.1%

Region VI

₱47.98B · 3%

Central Office

₱43.10B · 2.7%

Negros Island Region

₱38.60B · 2.4%

Region XII

₱35.87B · 2.3%

Region IX

₱35.25B · 2.2%

Flood-control contract value on the DPWH record, summed by the record's own region field across 18 regions (₱1.586T total). Statistical indicators derived from public data. Patterns may have legitimate explanations.

Corporate registry (SEC)

Ratio of a firm's total flood-control contract value on the DPWH record to its paid-up capital per its 2025 SEC GIS. Contract totals span 2021 to 2025; paid-up capital is the latest 2025 figure, which for firms that raised capital over time is the most conservative denominator. A statistical indicator only.
Firm	Paid-up capital	Flood-control value	Value : capital
M.G. Samidan Construction and Development Corporation	₱250,000	₱14.60B	58,386.5x
Centerways Construction and Development Inc.	₱45,000,000	₱11.97B	266x
Wawao Builders Corporation	₱50,000,000	₱8.40B	168.1x
Legacy Construction Corporation	₱380,000,000	₱25.72B	67.7x
Sunwest, Inc.	₱5,562,500,000	₱32.83B	5.9x

Shared registered office

Sunwest, Inc. and Hi-Tone Construction and Development Corporation register their principal office in the same locality: Lidong, Sto. Domingo, Albay.

Registered principal office in the same barangay/municipality per their 2025 SEC GIS. A recorded co-location indicator, not evidence of a shared operation.

Charges and flags are allegations under the presumption of innocence. These are corporate-registry facts and structural indicators, not findings of wrongdoing about any person or firm.

Predicted ties (statistical, unverified)

Node2Vec graph embedding (64 dims, 10 walks of length 30 per node, seed 42) over the flood-control firm-to-district-office graph plus recorded joint-venture edges. For firm pairs with >= P1B recorded flood-control value each and NO recorded joint venture, the score is the cosine similarity of their embeddings; Adamic-Adar over shared offices is reported alongside for corroboration.

Firm pair (no recorded joint venture)	Score	Offices	A-A
RCDS CONSTRUCTION & SUPPLY, INC.(FOR RCDS CONST + WEARR ENTERPRISES & DEVELOPMENT CORP	0.949	5	1.90
M3 KONSTRACT CORPORATION + 729 CONSTRUCTION, INC	0.945	6	1.83
3DIAMONDS CONSTRUCTION CORPORATION + WEARR ENTERPRISES & DEVELOPMENT CORP	0.943	5	1.79
JHI CONSTRUCTION & SUPPLY + W.J.P. CONSTRUCTION AND SUPPLY	0.939	3	0.73
RS RODRIGUEZ CONSTRUCTION & SUPPLIES + W.J.P. CONSTRUCTION AND SUPPLY	0.936	3	0.73
KENWOOD CONSTRUCTION ENTERPRISES CORPORATION + JEUSMACK BUILDERS & CONSTRUCTION SUPPLY	0.933	2	0.52
I.A . BOSQUE CONSTRUCTION CORPORATION + ROMHEAD S.J. BUILDERS AND DEVELOPMENT INCORPORATED	0.929	2	0.63
BET CONSTRUCTION & SUPPLY + J.G. GARCIA CONSTRUCTION & SUPPLY (FOR: J.G. GARCIA CONST.)	0.921	4	0.99
GRAIA CONSTRUCTION CORPORATION + KLDS TRADING & CONSTRUCTION	0.921	3	0.72
SYMS CONSTRUCTION TRADING + A.R.L. CONSTRUCTION	0.920	1	0.20

A predicted tie is a statistical similarity in bidding footprint. It is NOT evidence of a relationship, coordination, or wrongdoing, and it has not been verified against any registry or filing. In the explorer, these render as the faintest tier and are off until switched on in the legend.

How this was built, and its limits

Methodology

What this is, and is not

A map of public records. Every node and edge is either a recorded fact from the DPWH transparency dataset or a curated, source-linked action (a license revocation, a blacklist, a court filing, an audit flag). The graph computes descriptive statistics, concentration, co-award structure, who won the most. It does not establish guilt, and it is not a list of wrongdoers. Charges are allegations under the presumption of innocence.

Data sources

DPWH contracts. 248,220 contracts worth ₱6.379T, from the DPWH Transparency Portal via BetterGov.PH HuggingFace dataset, CC0 . License: CC0 1.0 Universal (public domain).
Official actions. Curated and source-linked: PCAB Board Resolution 075, Ombudsman and Sandiganbayan filings, DPWH Secretary orders, COA Fraud Audit Reports (attached to the contractors and the procuring district office they name, with the specific project, amount, and finding), SEC resolutions.
News tagging. Recent coverage matched by exact firm name to the firms named in the flood-control reporting, each linking its source article.

Entity resolution

The DPWH contractor field is messy: the same firm appears with and without its numeric id, inside joint ventures (split on "/"), and with "formerly" and "[REVOKED]" notes. We resolve each firm to its DPWH id (or a normalized name when no id is present), collapsing 248,220 contract rows to 6,558 distinct firms. Joint ventures become recorded co-award edges. For example, the 3,902 contracts tagged "[REVOKED]" resolve to just 8 firms (₱182.04B). A naive count of the raw strings would report 256.

Confidence tiers

An inferred link never looks like a recorded one.

Recorded (solid line): a contract award, joint venture, revoked license, blacklist, court filing, a source-linked person, or an SEC corporate-registry fact from a firm's own General Information Sheet, with a source.
Inferred from records (curved, lighter): not stated but computed, such as two firms that are both top awardees in the same district offices.
Predicted (faintest, off by default): a Node2Vec statistical similarity in bidding footprint between firms with no recorded joint venture. Not evidence of a relationship, and unverified against any registry.
Possible namesake: a shared surname is not a relationship. Not shown in this release; reserved for a future human-verified layer.

SEC corporate registry

Philippine SEC company data is not machine-readable in bulk: eSEARCH is a paid, authenticated document channel, automated scraping is not authorized under the Revised Corporation Code, and the SEC API Marketplace is subscription-gated. So this layer is not scraped. It curates the primary SEC documents, General Information Sheets and Articles of Incorporation, that PCIJ obtained and published for 6 of the top flood-control contractors PCIJ: Records of flood-control contractors . Each figure (registration number, paid-up capital, registered office) is transcribed from the firm's own GIS and links back to that document. The contract-to-capital ratio is then computed from that paid-up capital and the flood-control value already on the DPWH record, so the numbers reconcile. Where a figure was not legible in the published document, it is omitted, not estimated.

Geography and jurisdiction

Flood-control value is mapped to jurisdictions using the DPWH record's own region field, summed across 18 regions (₱1.586T total, reconciling to the site figure). Two joins are deferred, not faked: canonical PSGC province and municipality codes (the PSA publishes PSGC only as manual downloads behind Cloudflare, so there is no automated join), and the Open Congress bill layer (the documented BetterGov Open Congress API returned only a 404 page at every path this cycle). A point-level project map is a natural next step; the record carries coordinates on most contracts.

The numbers

Node size is not an importance score, but the recorded flood-control contract value. Concentration uses the Herfindahl-Hirschman Index per district office; 20 flood-control offices exceed 2,500 (the US DOJ "highly concentrated" threshold). Brokerage (betweenness), influence (PageRank), and co-award communities (Louvain, 15 groups) are computed offline with networkx and baked into the graph. Flood control is pinned to the exact DPWH category, 33,866 contracts worth ₱1.586T.

Network analysis methods

Computed offline with networkx and gensim, seeded for reproducibility, and baked as JSON. Temporal: per-year value, the named firms' share (contract value split equally among joint awardees so shares sum to 100%), newcomer share, and the cumulative joint-venture network. Patterns: descriptive indicators with stated thresholds, in the spirit of the OECD guidelines for fighting bid rigging in public procurement; a zero result is reported as tested. Predicted ties: Node2Vec embeddings (64 dims, seed 42) over the firm-office graph; cosine similarity between firms with no recorded joint venture, corroborated by Adamic-Adar. Statistical similarity only, never presented as a relationship.

Temporal knowledge graph

The graph is also read as a time series of facts. Temporal link prediction trains on the awards up to each year and predicts next year's new joint ventures from prior shared-office structure (time-decayed Adamic-Adar), on a strict chronological split, validated against both a label-permutation test and a degree-preserving configuration-model null. Dynamic community detection runs Louvain per year with adjusted-Rand stability across years. Change-points use Pettitt's non-parametric test on each structural series. Temporal motifs compare a joint venture's year to the median year of the pair's shared awards. A heterogeneous schema (person, firm, office, institution nodes; typed dated edges) holds the money-and-power layer, populated only from sourced records.

Verification and limits

Primary-source-or-omit: an official action or person enters the graph only if it traces to a primary or primary-citing source. Firms without a confirmed action carry recorded facts only. The person layer is curated and source-linked, not scraped. Not yet joined, stated plainly: bulk SALN wealth and SOCE campaign-finance records (neither is available as machine-readable public data today; the one campaign-finance link shown is individually sourced), and the PhilGEPS cross-check. Oversight bodies: Independent Commission for Infrastructure (ICI), Senate Blue Ribbon Committee, Office of the Ombudsman, Commission on Audit (COA), Anti-Money Laundering Council (AMLC), Department of Justice (DOJ).

All data sourced from public records (COA, SEC, DBM, PSA, BSP, SALN disclosures). This tool computes statistical indicators only. Specific allegations, if any, require independent investigation and corroboration.

Related projects

BetterGov.PH flood visualizations Rappler Politicontractors InfraWatch PH PCIJ MoneyPolitics OCCRP Aleph LittleSis