NEP and final GAA provide the published bookends. The investigative focus is how allocations change through House and Senate deliberations and bicameral reconciliation. A genuinely new HGAB line deserves scrutiny: who proposed it, what documented need supports it, and whether it remains in the Senate and reconciled versions. A lawful amendment can add projects; an unmatched parser row is not yet a demonstrated amendment.
These pages currently trace NEP and House readings. No survival claim is made for Senate GAB, Bicam or FY2027 GAA until those versions are available and reconciled. Retained NEP repeat candidates and new HGAB insertion candidates answer different questions.
What policy or engineering design justifies one identical budget unit per office?
Observed evidence
All district-office RCS lines have the same amount. Province/regional rainfall proxies differ by 2.59×. Alternative formulas move different numbers of budget units. All 199 DEO lines are retained unchanged in HGAB; no differentiated needs weighting is visible in the amounts.
Plausible explanation
Standard package costs or a minimum coverage policy may justify equal allocations. Rainfall alone does not measure demand.
What investigators need next
Obtain siting, design capacity, water demand, catchment area and existing RCS inventory.
Parsing/location caution
DEO is not LEG; province rainfall is not a district observation. Returning FY2026 program.
5,549 supplied-comparison rows vs 6,105 / 15,979 native project rows
Talking point
Which source rows explain the difference between the supplied comparison and the native one-to-one screen?
Observed evidence
The supplied comparison marks 5,549 domestic rows without a plausible NEP title counterpart, carrying ₱158.363B. The native one-to-one screen selects no NEP counterpart for 6,105 of 15,979 HGAB project rows. The 556-row difference is not a count of missing projects: the screens use different matching rules and row populations.
Plausible explanation
Different project filters, grouped occurrences, parser boundaries, source versions and matching rules can change candidate counts.
What investigators need next
Reconcile project IDs and source pages, align project-level denominators, then adjudicate a sample against the original NEP and HGAB pages.
Parsing/location caution
Neither count is a certified insertion total, and the counts are not independent confirmations of the same set of projects.
Which MPB lines are genuinely new, and what documented facility needs and designs support them?
Observed evidence
The HGAB MPB PAP contains 5,765 lines carrying ₱83.707B. 620 counterpart cases remain separate from insertion candidates. The historical screen has 819 comparison pairs covering 587 distinct NEP proposals.
Plausible explanation
Renaming, combined occurrences and legitimate phased construction can explain unmatched or recurring titles. Similar amounts can reflect standard designs.
What investigators need next
Audit original NEP/HGAB pages; obtain building purpose, floor area, design, quantities, contract scope and completion records.
Parsing/location caution
MPB overlaps the full DPWH insertion/repeat findings; do not add their amounts. Equal allocations are not equal construction unit prices.
1855 of 2135 NEP comparison pairs clearly retained in HGAB
Talking point
Are new appropriations paying for already-funded components or legitimate continuation?
Observed evidence
1194 of 1379 distinct NEP repeat-screen proposals have clear HGAB links. 24 proposals have explicit overlapping historical chainage. ₱48.207B is the unique NEP proposal allocation, not alleged loss.
Plausible explanation
Multi-year implementation, rehabilitation and distinct components can share names and segments.
What investigators need next
Prioritize explicitly overlapping retained proposals and high amounts; compare contracts, quantities, tranche scopes and completion.
Parsing/location caution
The 737 headline counts NEP–old GAA comparison pairs. Potential retention and missing counterparts remain separate.
Can the funded segment and engineering length be established for every allocation?
Observed evidence
Roads: 3,329/6,482 lines have derivable stated lengths · Bridges: 65/1,159 lines have derivable stated lengths · FMR: 3/964 lines have derivable stated lengths. Official programs have separate denominators on the supporting page.
Plausible explanation
Lengths may be recorded in engineering schedules or continuation text not recovered by this extraction; incomplete metadata need not reflect deliberate omission.
What investigators need next
Inspect original source pages and obtain funded-segment boundaries, complete chainage, project quantities and bridge spans; reconcile extraction gaps before alleging disclosure failures.
Parsing/location caution
Missing includes absent and unusable extracted evidence, not proof the original documents lack length. Coordinate distances do not establish construction length; mixed scopes are separate.
Do apparently high rates reflect short printed segments, structures, broader funded scope, or unusually costly comparable work?
Observed evidence
Whole HGAB line-item allocation is divided by its derivable stated segment meters. Roads, bridges and FMR are displayed separately, with official work-program filters. Mixed road/bridge lines receive no single-work rate. Data-derived length groups and within-program slope-change screens are provided; no fixed 100-meter boundary is imposed. Lane evidence now supports 3 title-stated lane rates, 279 stated-length inventory lane estimates, and 390 existing-bridge asset lane estimates. These bases are shown separately and are not pooled into the per-meter flags. Recorded-condition screening identifies 2 rehabilitation lines on fully matched Good/Fair assessed road segments and 0 pavement-work lines within the disclosed five-budget-year surfacing window. These are review leads; later deterioration and different work scope may justify the allocations.
Plausible explanation
Terrain, width, structures, approaches, rehabilitation scope and components outside the stated segment can explain large differences. Tiny printed ranges or parser errors can produce extreme ratios.
What investigators need next
Verify station endpoints on original pages, then compare quantities, width, funded extent, designs and bill-of-quantity costs within the same work type. Investigate unusually short denominators before interpreting extreme ratios.
Parsing/location caution
These are appropriations per stated segment meter, not verified construction unit costs. Bridge chainage is not necessarily bridge span. No cross-category average, overpricing estimate or corruption probability is established. Inventory lanes are existing asset characteristics, not verified funded quantities; widening needs additional funded lanes. Overpasses, flyovers and viaducts use bridge peers.
Do the printed station endpoints correctly describe the funded segment, or do notation and scope need correction?
Observed evidence
Original-page visible text agrees with sampled road titles whose station differences calculate to 0.587 meters, 1.49 meters, 4.67 meters. Across all 3,175 known road lengths, only 204 (6.43%) are below 100 meters; high-rate sorting concentrates short denominators at the top.
Plausible explanation
A source-document typo, station convention, short component within a larger project, or broader funded scope can explain these values. A very short printed range does not establish the physical length actually funded.
What investigators need next
Request corrected or certified station endpoints, funded-segment plans, engineering quantities and allocation scope; verify notation before interpreting extreme allocation-per-meter rates.
Parsing/location caution
This is a printed-range/denominator issue, not evidence that most road projects are short or overpriced. Sampled title agreement rules out a title-transcription mismatch for those examples, not source errors or engineering ambiguity. Bridge lengths require separate interpretation.
Which HGAB 3rd projects are genuinely absent from NEP after title, scope and occurrence review?
Observed evidence
The current one-to-one comparison selected no NEP counterpart for these project rows. The comparison uses title and scope signals; unresolved and unmatched cases are not proof of new projects.
Plausible explanation
Renaming, title wrapping, scope changes, occurrence aggregation and parser errors can prevent counterpart selection.
What investigators need next
Review NEP and HGAB source pages side by side, then reconcile project scope, office, location and amount.
Parsing/location caution
No selected counterpart means a review candidate, not a confirmed insertion.
Which listed attachment types are absent, and can DPWH provide the missing POW, photos or RDC?
Observed evidence
Counts cover only title-matched HGAB projects whose attachment metadata is locally inspected. The supporting section breaks missing POW, photos and RDC out separately.
Plausible explanation
The local snapshot may be incomplete, and an attachment may be indexed under another type or record.
What investigators need next
Review each project’s transparency entry and request missing categories from the implementing office.
Parsing/location caution
Uninspected projects are unknown, not counted as missing; the missing-type groups can overlap.
NEP → HGAB (House) → GAB (Senate) → Bicam → GAA are different stages; prior-year GAA is enacted, current proposals are not. Match allocation records at the same hierarchy before comparing money.
Printed-total reconciliation catches amount/hierarchy errors but cannot certify title identity or location.
Preserve source pages and occurrence counts. One comparison row may group multiple allocations; headers are not projects.
Require barangay → town/city → province context. DEO numbers do not identify legislative districts.
Keep directional names, tranches, province splits and construction chainage distinct. The confirmed multi-place Butuan/RTR line stays outside strict LEG totals.
Strict congressional rankings cover only 41.4% of extracted allocation value. Names indicate geography, not authorship or sponsorship.
Compare the same budget hierarchy before calculating changes. Missing comparable NEP data is not zero. The schedules and matched subsets below overlap other findings and must not be summed.
Scope
NEP allocation pesos
HGAB allocation pesos
Allocation adjustment pesos
Coverage limit
FMR annex
₱16,000,000,000
₱19,628,100,000
₱3,628,100,000
Programme controls; individual counterpart confidence stays separate
NIA exact named-title subset
₱10,355,321,000
₱8,855,321,000
₱-1,500,000,000
Matched subset only; not the whole NIA budget
HFEP programme total
₱14,539,832,000
₱10,088,832,000
₱-4,451,000,000
MOOE and Capital Outlays matched at programme level
NIA irrigation annex
₱17,598,181,000
₱17,999,305,000
₱401,124,000
Reviewed detailed annex controls; project identity remains separate
DepEd school operations
Not established
₱102,670,874,000
Not established
Comparable NEP school baseline not built
Evidence needed to strengthen the claims
Investigative question
Dataset or documents needed
Current workspace limitation
RCS needs
Geocoded RCS sites, storage/catchment design, population served, water access and dry-season demand
Rainfall is province-level; it cannot locate demand or actual sites
Repeated work
Contract IDs, bill of quantities, as-built chainage, completion and disbursements
Budget titles and stated chainage do not establish paid work
Price clustering
Work quantities, standard estimates, procurement thresholds, prior-year comparable contracts
Exact peso values alone are not unit prices
Insertions
Audited native NEP/HGAB occurrence links and original PDF pages
A balanced total does not certify every unmatched row
LEG rankings
Dated barangay/municipality LEG boundaries and funded-segment geometry
PH location DB contains undated and conflicting claims
INSPIRE
Approved total project cost, phase ledger, obligations and releases
Appropriations are not disbursements
Workspace search found a province poverty file, but its rows do not identify year or indicator definition; it is not used as a needs control. The local flood project summary covers 2016–2024 projects, not present-day hazard exposure. Neither provides the missing RCS site/catchment, water-access or demand denominator. Additional datasets must be dated, sourced and matched to the correct geographic level before correlation.
Full line-item review queue
One row per domestic DPWH HGAB 3rd-reading allocation occurrence. The full queue contains 16,083 flagged rows, sorted by flag count, allocation amount, then occurrence ID. The chart shows the first 10; the paginated table below covers the entire queue. Up to twelve review flags cover amount patterns, NEP counterparts, listed documents, POW/photos/RDC coverage, historical recurrence, chainage, rate, recorded condition, recent surfacing and very short printed ranges. Each icon marks a reason for a human to check the evidence. Icons do not represent findings or levels of guilt. Document-type flags apply only to locally inspected attachment metadata; unavailable or inapplicable evidence is not treated as verified.
Icon key: 📐 short range · 📈 relative rate · 🛣️ recorded condition · 🛞 recent surfacing · 🧩 NEP match review · 📂 listed documents · 🧮 POW · 📷 geotagged photos · 🏛️ RDC · 🪙 repeated amount · 🕰️ historical title · 📍 chainage overlap
Top 10 projects in queue order · flag count, then allocation
Loading ranked rows…
Rank
HGAB project
Program
Implementing office
Flags for human review
HGAB allocation pesos
HGAB source
How the ranking is calculated
Potential insertion: the current matcher identifies no plausible NEP counterpart; ambiguous links and FAPs are excluded.
No listed documents: no matched DPWH Transparency entry with listed documents exists in the current snapshot. This is independent of NEP counterpart status; documents may exist elsewhere.
Incomplete documentation: each absent POW, geotagged-photo or RDC type earns one point only when that project attachment list was locally inspected. No type is inferred missing for uninspected projects.
Repeated allocation amount: the same exact positive amount occurs in the same PAP at least 10 times across at least 3 named implementing offices. Roundness alone adds no point.
Historical title recurrence: an FY2020–FY2026 DPWH GAA title agrees after case/punctuation normalization and removal of complete trailing station/coordinate metadata, with qualified municipality and province agreement. FY2020–FY2025 use ODV annual GAA data; FY2026 uses the DPWH Details Enrolled Copy. Work descriptions, directions, packages and phases remain; conflicting locations/tranches and disjoint chainages are excluded.
Historical chainage overlap: an accepted historical counterpart has explicit, positive-length overlap. This is a scope indicator, not proof of duplicate paid work.
Very short range: a recovered printed project length under 10 meters earns a review flag. Verify notation and funded extent; it does not establish an unreasonable allocation.
Each qualifying type contributes one point, up to twelve possible types. Multiple historical years do not increase points. Historical recurrence and chainage overlap are related signals, so the count is not a statistical probability or independent-evidence score. Missing historical links can reflect conservative matching, incomplete coverage or unresolved names. Standard packages and legitimate amendments can produce these signals. Amounts overlap other findings; do not add them as alleged losses. Names, roles and party affiliations add no points. The complete review queue is also available as full ranking JSON.
Condition and prior-surfacing flags
One point for rehabilitation/reconstruction/major repair when the full matched road range was assessed Good/Fair in a known pre-budget year. One point for pavement work when the full matched range has a positive recorded last-surfacing year within five budget years. Widening, approaches, slope/drainage and off-carriageway scope are excluded. Unknowns and partial coverage add no point. The five-year window is a screening choice, not an engineering service-life standard. Bridge Good/Fair context without an assessment date remains a review lead and earns no dated-condition point. Later deterioration or different funded components may justify the work. Condition, recent surfacing and historical recurrence can be related; points are not independent evidence or probabilities.
Cost-per-meter and short-range flags
One point when allocation per stated meter exceeds both twice the peer median and Q3 + 1.5 × IQR. Peers share the work category, structure group, matched surface/condition/assessment-year context, exact official PAP and a data-derived length band, with at least 20 records whose stated lengths are at least 10 meters. No fixed 100-meter cutoff is used. Separately, a recovered printed length under 10 meters earns a denominator-warning point; verify the notation and funded extent. Extreme rates remain in this review table with a caution but are excluded from rate charts. Missing lengths and insufficient peers add no point. These flags do not establish overpricing: width, terrain, structures, quantities and funded extent remain unverified. See full rate and grouping methodology.
How all twelve summary findings relate to this queue
Summary finding
Treatment in the line-item ranking
1 · Equal RCS packages
Repeated-amount flag if its existing thresholds qualify; no extra point for belonging to RCS. Needs weighting requires installation evidence.
The external NIA House annex contains 828 project rows / PHP 17,999,305,000; the local detailed schedule contains 828 / PHP 17,999,305,000. The external NEP detailed annex is now available (791 rows). The reviewed extraction takes priority over older local automated parsing. Newly recovered rows are parser corrections, not budget insertions.
The external retained NEP/House comparison uses the SECOND reading and reports 5801 unmatched House candidates and 1341 unmatched NEP candidates. Our principal analysis uses the THIRD reading. Chainage/re-scoping fallbacks consume some external pairs; differing stations and altered geography still require independent review. The earlier 5,837/2,139 figures are not a current target.