Methodology · v2.9

RETTransparence du lobbying

Lobbying / regulatory-engagement disclosure completeness for {legal_name} audited across four dimensions: (a) registration completeness on EU Transparency Register / US Senate LDA / national-equivalent registers (UK OSCR / DE Lobbyregister / FR HATVP / IT Rappresentanti di interessi presso il Consiglio di Stato), (b) spend disclosure granularity, (c) named-lobbyist disclosure, (d) policy-position publication. ENTITY-MATCH RULE (mandatory, anti-confusion): every cited register entry / lobbying filing MUST belong to {legal_name} ITSELF (matched by legal name + corporate domain + jurisdiction) or to a named industry association it demonstrably belongs to. REJECT same-name-but-unrelated entities — a lobbying/consulting/law firm that merely has a similar name (e.g. a firm whose name happens to contain the company's name), a same-name person, or an unrelated registrant in a jurisdiction where the company has no presence. Citing a register page WITHOUT a confirmed {legal_name}-specific entry (e.g. a bare state/EU register homepage with no matching filing) is a protocol violation — that sub-signal is then 0/absent, not a guess. If NO genuine {legal_name}-specific lobbying disclosure can be found AND no industry-association membership can be confirmed (see proxy below), return absent{reason:"not_disclosed"} rather than scoring on a wrong or unverified entity. INDUSTRY-ASSOCIATION PROXY (MANDATORY check BEFORE bailing — regulated sectors such as banks/insurance/pharma/auto/energy/utilities/infrastructure rarely register directly and almost always engage via associations): BEFORE returning absent you MUST check whether {legal_name} engages through a recognized industry/trade association (e.g. ABI / Federbanche for IT banks, ABA / SIFMA for US banks, EBF for EU banks, Insurance Europe / ANIA for insurers, ACEA / VDA / ANFIA for automakers, PhRMA / EFPIA / Farmindustria for pharma, Eurelectric / Utilitalia / Confindustria-Energia for energy & utilities, etc.). EVIDENCE BAR (deliberately light): EITHER (i) {legal_name}'s own disclosure of membership (its website / annual or sustainability report naming the association) OR (ii) the association's member directory listing {legal_name} is SUFFICIENT to score registration_completeness as PARTIAL (typically 35-55) at confidence=preliminary — you do NOT also need the association's separate lobbying registration. The remaining sub-signals (spend / named-lobbyist / policy-position) are scored on whatever direct disclosure exists, else low. This prevents absent for regulated companies whose engagement is almost entirely association-mediated. ANTI-FLOOR (mandatory): if a dimension genuinely DOES NOT APPLY (e.g. the company engages ONLY via association and does not lobby directly, so spend / named-lobbyist / policy-position have no basis to disclose), OMIT that sub-signal entirely — do NOT return it as 0. composite_signal_score averages ONLY the sub-signals you return; emitting a 0 for a non-applicable dimension penalizes the company for not doing something it has no obligation to do (e.g. an association-mediated firm scoring ~14 instead of ~45). Return ONLY the applicable sub-signals; if NONE applies, return absent{reason:"not_disclosed"}. EXTRACTION RULES: Return composite_signal with one sub-signal per dimension (name = dimension label + value 0-100 reflecting completeness for that dimension + source_url backing it via EU Transparency Register / OpenSecrets / LobbyFacts.eu / US Senate LDA / national lobby registers / industry-association memberships). All 4 sub-signals recommended; minimum 3 for confident. Server applies composite_signal_score (weighted mean) — DO NOT compute any 0-100 score.