PK1/H1 granular procurement validation - FINAL

Universe: tenders.db, 61 local-government procurement pages.
Years: 2015-2024 loaded; inferential windows require >50% of municipalities and complete municipality coverage in every year of the window.
Missing municipality-year is never coded as zero. A zero sector award is used only when the municipality is observed in all years of the window but no awarded physical public-work procedure matches the target sector.
Main amount: winning offer excluding VAT, restricted to non-cancelled procedures, deflated to 2024 prices using INSTAT annual-average CPI growth.
Robustness amount: fund limit, only for procedures that reached an award and have a winning offer.
Public works: physical works only; design/supervision-only titles are removed.
Sector classification: conservative title-based rules; a project may have multiple labels when it explicitly contains multiple sector components. Strict robustness retains only single-label target projects.
Territory: database municipality except when contracting institution identifies another municipality directly, or a regional utility/county council tender title uniquely identifies the beneficiary municipality.
Need: authoritative Need_Snapshots_Final from PK1_H1_Analiza_PERFUNDIMTARE_AUDITUAR.xlsx.
Relative/rank need is compared with matched award per snapshot-year population; absolute burden with absolute matched award.
Specificity: rho(problem, matched sector) - rho(problem, public works without the target sector), with 3,000 municipality bootstraps.
Scale: log1p(matched award) on log1p(absolute burden), ln(population), ln(area), HC3.
BH-FDR: within family x need dimension x H1 role, separately for each outcome family.
2011 water/sanitation procurement tests are explicitly delayed validation starting in 2016; they do not recover missing 2012-2015 procurement and are not treated as immediate t+1 tests.