Cost per meter · stated-length coverage

Which projects have unusually high allocation per meter?

The first tables compare each rate with the typical rate in the same official program and a similar stated-length group. A high-rate candidate is more than twice the typical peer rate and unusually high within its peer group. This is a scope-review queue, not a verified overpricing finding. Extreme values are omitted from rate graphs and described alongside them; they remain in the tables and CSV. Coverage denominators include all selected extracted occurrences in each category, including those without a usable length.

These are whole line-item allocations divided by stated segment lengths, not verified construction unit costs. Bridge chainage may cover approaches rather than bridge span. Missing percentage means no usable length was recovered from the current title/project context; it does not establish that the original budget or engineering documents omit length.

Download all records and length evidence

Are stated-meter allocations higher than FY2026 GAA?

Compare the same official work program first, then comparable stated-length groups. FY2026 is enacted GAA; FY2027 is the HGAB 3rd-reading proposal. These are allocations per stated meter, not verified construction prices or a finding of overpricing.

Coverage and exclusions

Method and limits

Same length parser as current page; positive domestic DPWH OPERATIONS col_J occurrences only from local parsed FY2026 schedule. Peso amounts already in pesos. Official PAP joins use case/punctuation normalization only; unmatched programs excluded from peer comparisons. Lengths below 10m withheld from rate comparisons. Program medians need at least 20 usable records per stage. Length adjustment uses FY2026 length quartiles, at least 20 records per stage per band, and a FY2026-count-weighted mean of band median ratios. Unsupported bands are excluded, never extrapolated.

• Different projects and stage scope; not paired contracts or measured construction cost inflation.
• Width, lanes, terrain, structures, geography and funded quantities remain uncontrolled. Rates are whole allocation/stated meters.
• FY2026 column-J leaf scope can omit records at other hierarchy depths; selected coverage is not the complete DPWH budget. Missing metadata can bias comparisons.
• FMR is not compared here: this historical comparison reads the DPWH schedule only.
• Medians resist extreme-rate influence; no extreme individual rate dots or bars shown. No formal significance or causal claim.

Download all FY2026 records, band calculations and input hashes

Supporting length-group calculations

The adjusted percentage is the mean of the group rate ratios, weighted by the FY2026 record count. Only the groups shown qualify; omitted groups do not count toward the result.

Very short printed ranges · original PDF text checks

The following extracted titles agree with visible text on the cited original PDF pages. This establishes printed notation, not its intended engineering meaning or the actual funded length. In particular, sub-meter road ranges require endpoint and scope clarification before interpreting allocation rates.

Download source-check evidence and full-population counts. Tiny printed ranges may reflect source notation or source-document errors. Bridges can legitimately be shorter than roads; a short bridge range alone is not an anomaly.

At what stated lengths does allocation density change?

Length bands are compared within the same category and official work program. Allocation density is pesos per stated meter. A shorter denominator can raise this ratio without an unusually large allocation, so the tables also show total allocation and a separate model of allocation versus length.

These are exploratory grouping boundaries, not engineering price standards or evidence of overpricing. Sparse samples, missing lengths and unmeasured project scope limit interpretation.

Where median density changes by at least 2×

Only comparisons with at least 20 rows in each band are included. A ratio above 1 means the shorter-length band has higher allocation per meter. Bands with fewer observations remain in the complete band tables below.

Does total allocation show a length boundary too?

The segmented model fits log total allocation against log stated length within each PAP. It needs 80 records and 20 per side, excludes lengths below 10 meters, and requires BIC improvement ≥6 over a single slope. The boundary is selected from observed-length candidates, not a fixed 100-meter threshold; bootstrap frequency describes how consistently it is selected. The 10th–90th percentile of bootstrap-selected boundaries describes an empirical stability window, not a confidence interval. The ≥25-meter sensitivity boundary is also shown.

Complete length bands · separate categories and programs

These descriptive medians remain separate by PAP. FMR and bridge samples may be too sparse for breakpoint inference.

Method and limitations

Length bands are derived separately within each PAP using approximately equal-count groups (up to eight, targeting at least 20 records per group). Equal observed lengths stay together. Breakpoint candidates are geometric midpoints between adjacent observed distinct lengths, requiring at least 20 records on either side; up to 31 candidates cover the eligible boundaries. No 100-meter boundary is prescribed. The searched boundary counts as an extra model parameter in BIC.

• Density=allocation/length; declining density can follow mechanically from a fixed budget.
• Quantile-band counts and 2×/BIC screens are editorial exploratory choices, not calibrated corruption thresholds.
• Models use total allocation rather than density to reduce the arithmetic-denominator trap.
• Bootstrap votes measure selection stability in this sample, not the probability of a real engineering threshold.
• Dependence between allocation occurrences, unknown width/terrain/structures and non-random missing lengths remain uncontrolled.
• Sub-10m rows stay visible in bands but are excluded from model fits; ≥25m sensitivity is reported.
• Multiple exploratory program/boundary searches are not confirmatory significance tests; no p-values or multiplicity-adjusted claims are made.
• Sparse bridge/FMR data cannot establish an engineering threshold. No pooled cross-category rate model.

Download all groups, model slopes, bootstrap votes and methodology. Patterns direct source and engineering review; there is no universal safe or suspicious length.

Length extraction and coverage methodology

• Complete STA/K/KM plus-form ranges are converted to meters. Overlapping numerical intervals are merged, never double-counted. Multiple named routes with unverified station references are withheld.
• A single explicitly labeled length in meters or kilometers can supply a length. Widths, isolated stations and coordinate straight-line distances do not supply a road length.
• FMR continuation lines use only the existing parser's same-row project-column context. Context preceding the current project title is discarded, preventing a previous project’s endpoint from becoming part of the current length. Exactly two printed STA endpoints in the remaining context can form a range. The context is retained in the downloadable data; nearby unrelated text is not searched for endpoints.
• Partial or zero-length chainage, conflicting explicit lengths and ambiguous multi-route scopes remain without a usable length. No allocation amount is used to infer missing length.
• The missing percentage is (all category occurrences minus occurrences with usable stated length) / all category occurrences. It includes both absent metadata and metadata that needs review. A separate rate count distinguishes usable lengths from permissible single-category ratios.
• Mixed road/bridge titles are separated and no single-work rate is shown. Flood, building/water PAPs and foreign-assisted schedules are outside this page's denominator. Classification uses explicit titles and PAPs; it is not a complete inventory of physical road assets.
• No rate averages are pooled across categories or PAPs. Every point identifies one allocation occurrence; rankings can change with scope, width, structures, terrain, total project components and parser corrections. High rates warrant source/quantity review, not an allegation.
• Graphs and CSV follow table filters. Paired bars compare project allocation/meter with the median of its own peer group; up to 12 non-extreme records are displayed. Extreme values are described beside the graph; all filtered rows, including missing lengths, remain in the table/CSV. Source volume/page and original length evidence are included. Historical stages are not used to fill current-stage gaps.

The summary links to two separate leads: length-data coverage and allocation per stated meter. Coverage gaps are extraction observations, and ratios are scope-dependent screening proxies. Extremely short printed ranges can produce extreme rates; verify station endpoints and funded scope on original pages before interpreting them. Request bills of quantities, designs, funded segment boundaries, bridge spans and roadway widths before comparing actual costs.