AN-095: VS Publicidade deep dive

VS Publicidade LTDA (the top-ranked AN-094 shell) routes 218 of its 254 mayoral polls (86%) through Publi. QC Pesquisas & Editoração — and Publi. QC self-contracted all 230 of its 2020 mayoral polls in São Paulo without an intermediary. The Goiás IPOP/FacUnicamps cycle-over-cycle shift recurs at this scale in São Paulo with a different cover entity. Owners are different people (Sanazar brothers for VS, Flávio Henrique da Silva for Publi. QC), so the relationship is contractual exclusivity rather than common ownership.

Hypothesis
H13: Shell-contratante polls show larger residual β
Confidence
green
Type
descriptive
Design
Sample
VS Publicidade LTDA (CNPJ 96499132000189), all 254 mayoral polls it sponsored in 2024, plus the 2020 mayoral-poll universe registered by the same pollster (Publi. QC Pesquisas & Editoração, CNPJ 37658984000102) for cross-cycle comparison
Specification
1. Pull all sponsor rows where sponsor_id = 96499132000189 from pipelines/politica/build/clean/poll_sponsor_2024.parquet. 2. Join to poll_2024.parquet for full registry metadata (institute, dates, sample size, value_brl, muni, UF). 3. Tabulate pollster concentration (which pollster did VS hire, how often), UF concentration, time distribution. 4. Cross-cycle: pull Publi. QC's 2020 mayoral polls from poll_sponsor_2020.parquet; compute sponsor structure (was Publi. QC self-contracting in 2020?). 5. Partner data: load May-2025 RFB socio snapshot (pipelines/cnpj/build/clean/socio_202505.parquet, 8-digit cnpj_base key) for both VS Publicidade and Publi. QC.
Comparator
Goiás IPOP/FacUnicamps cross-cycle pattern (paper §2)
Script
source/analysis/an-095-vs-publicidade-deep-dive.py
Target
build/intermediate/an-095-vs-publicidade.csv
Status
interpreted · 2026-06-17
Created
2026-06-17

Question

VS Publicidade LTDA was named in the §2 national-pattern paragraph added in AN-094 as the single largest residual-tier shell in 2024 (254 mayoral polls registered). The paragraph is currently grounded on top-25 enumeration alone. To make it a worked case alongside the Goiás FacUnicamps anecdote that §2 already has, we need three things: the firm itself (ownership, registered activity), the actual poll pattern (which pollster, which municipalities, which dates, which candidates), and whether the cycle-over-cycle shell-shift documented in Goiás also appears here.

Results

Firm

2024 poll footprint

Publi. QC, the paired pollster

Cross-cycle inversion

The Goiás case in §2 documents IPOP-Cidades shifting from self-contracting in 2020 (357 polls, all named IPOP itself as contratante) to FacUnicamps shell-contracting in 2024 (68 polls, all named FacUnicamps). The same shift occurs in São Paulo with Publi. QC and VS Publicidade — at very similar scale:

Pollster 2020 mayoral polls 2020 contratante 2024 mayoral polls 2024 contratante (218 of 219)
IPOP-Cidades (Goiás) 357 IPOP itself (self) 68 FacUnicamps (shell)
Publi. QC (São Paulo) 230 Publi. QC itself (self) 219 VS Publicidade (shell)

The cover vehicle changes across cycles; the pollster's volume does not. Both 2020 self-contracting patterns are exactly the disclosure-failure type Lei~9.504/97 Art.~33,~§3° formalizes as fineable but not criminal (paper §2). Both 2024 shell shifts move the same volume of polls through a third-party CNPJ whose web presence, declared activity, and ownership structure all fail to match the editorial role the registration assigns it.

Candidate side (partial — IPOP-iceberg layer present)

Interpretation

  1. VS Publicidade ↔ Publi. QC is the São Paulo analogue of FacUnicamps ↔ IPOP-Cidades in Goiás. Same cycle-over-cycle shift (self → shell), same single-pollster pairing, comparable scale (~220 polls / cycle), same registry signature (low or zero capital social, declared CNAEs the firm doesn't actually practice, no public web brand).

  2. The cross-cycle inversion is the load-bearing comparative fact. The paper's §2 closing line — "the cover vehicle changed across cycles; the pollster's volume did not" — is now exactly right for São Paulo, the country's largest electoral market.

  3. Different owners. Unlike a single-family enterprise, the Sanazar brothers (VS) and Flávio Henrique da Silva (Publi. QC) are unrelated. The exclusivity is contractual, not familial. This matters for the legal-asymmetry argument in §2: a same-owner arrangement could be characterized as integrated journalism; an arms-length contractual exclusivity at this scale (219 of 219 polls go through one sponsor) is structurally indistinguishable from a sponsorship-laundering arrangement, but Art.~33,~§3° reaches only the disclosure failure, not the contractual relationship.

Follow-ups

  1. Beneficiary identification (puzzle): which actual candidates' campaigns funded the 218 VS→Publi. QC polls? The 129 polls with response data show no single-candidate concentration — but per-poll, who was the favored candidate? A bias regression on the VS subsample alone (errror within race × week vs the subsample's leading candidate) could surface a candidate-level slant pattern even without funder identification. Suggested script: source/analysis/an-095a-vs-bias-by-race.py.

  2. VS Publicidade's 1 MS poll (blind spot): 253 of 254 in SP, 1 in MS. Worth checking which muni + which race; it may be a coding artifact, may be a one-off, may indicate the shell was tested cross-state.

  3. The 36 non-Publi.QC polls in VS's 2024 portfolio (puzzle): Quality Pesquisas (15), Vitoria Comunicação (10), Instituto Nova S.P. (5) — are these other small SP pollsters that VS Publicidade uses opportunistically, or do they show the same self→shell shift if you look at their 2020 cycle?

  4. MAC CLEAN SOLUÇÕES (blind spot): the 1 non-VS poll that Publi. QC did in 2024 was sponsored by "MAC CLEAN SOLUÇÕES E APOIO ADMINISTRATIVO LTDA" — a cleaning-services LTDA. Is that another shell in the same network, or a one-off legitimate contract?

  5. Were the Sanazar brothers ever candidates? (diagnostic): if either Marcelo or Marcio Sanazar appears in TSE candidato as a 2024 mayoral candidate, VS Publicidade is much closer to a Route A self-sponsored case than to a shell-cover one. Quick check against candidato.csv. Resolved 2026-06-17. Neither Marcelo Vrejhi Sanazar nor Marcio Rodrigo Sanazar appears as a candidate in any year. The only Sanazars in TSE candidato are Hirant Sanazar (Dep. Estadual 1998 PRONA / Dep. Federal 2002 PL, SP) and Sandra Regina Sanazar (Vereador 2008 PP, SP, 65 votes) — possible family connection to the VS partners, but neither is a 2024 mayoral candidate. The shell-not-Route-A characterization in §2 stands.

Caveats