The evidence · per-firm event-study battery

Did the market price the Texas move?

Bottom line. The per-firm event-study battery for MercadoLibre is pending the canonical run. The framework below lists the nine figures that will appear once MercadoLibre's daily-close panel (T−260 through T+10 around the canonical event date) is staged through the canonical pipeline. The methodology is invariant across firms; the data is firm-specific. See /methodology for the estimation design and inference framework.

The pre-registered event-window choice

MercadoLibre's canonical event date is a pre-registered choice between two candidates: the announcement date (Musk's X statement, 2025-06-09) and the effective date (8-K cover-page state change, 2025-06-10). The intermediate dates (DEF 14A April 29, 2024; meeting [DATA PENDING]; 8-K meeting results June 14, 2024) are not candidates for the canonical event because they are anticipatable from the announcement. The pre-registration document in 06_PREREGISTRATION/ documents the choice; pending publication of that document the event-date field carries [DATA PENDING].

The nine figures (pending per-firm run)

Each figure ships with a navy "Sources & methodology" strip with KaTeX-rendered formulas and a "View methodology" modal. The canonical XOM template is the visual reference; MercadoLibre's data populates each figure once the per-firm run completes. The figure list:

Figure 1 — Peer comparison. MercadoLibre's Day-0 AR alongside Latin American e-commerce peers (Amazon, eBay, Sea Limited, JD.com, Coupang, Alibaba) under the FF6 benchmark. [DATA PENDING: Day-0 ARs for MercadoLibre + peer set]
Figure 2 — Synthetic-control gap. ADH synthetic MercadoLibre against a 20-firm donor pool of large-cap consumer-discretionary firms. [DATA PENDING: donor weights + gap series]
Figure 3 — Cross-firm placebo. MercadoLibre's |gap| rank among a 21-firm placebo universe under the canonical 250-day pre-window. [DATA PENDING: rank + p-value]
Figure 4 — In-time placebo. 100 pseudo-event-date replications; MercadoLibre's actual gap against the placebo distribution. [DATA PENDING: placebo histogram + MercadoLibre position]
Figure 5 — Co-movement diagnostic. MercadoLibre daily returns vs. cohort daily returns over [T−30, T+30]. [DATA PENDING: paired-return series]
Figure 6 — Donor weights. SLSQP-simplex weights on the 20-firm donor pool; sum to 1.000000000. [DATA PENDING: weight vector]
Figure 7 — Three benchmarks. Synthetic control + matched pair + FF6 model, all three benchmarks reported with the bounded-null credibility band. [DATA PENDING: three Day-0 estimates + CI]
Figure 8 — Bayesian posterior. Conjugate normal-normal posterior on the true Day-0 effect; 95% credible interval; P(effect < −2pp). [DATA PENDING: posterior parameters]
Figure 9 — TSR vs cohort. Total shareholder return from MercadoLibre's announcement date through current; cohort-average overlay. [DATA PENDING: TSR series]

The 18-cell core battery (pending per-firm run)

Three benchmarks (FF6+BNO canonical, FF6 sensitivity, matched-pair consilience) × six windows ([0], [−1,+1], [−2,+2], [−5,+5], [−10,+0] leakage, [0,+10] outcome) = 18 cells. Each cell reports point estimate, Patell-z p-value, BH-corrected p-value, Romano-Wolf step-down p-value. MercadoLibre's 18-cell results: [DATA PENDING: per-firm run].

The bounded-null verdict (pending per-firm run)

The XOM canonical reports headline verdict "NULL across all 54 tests"; the bounded-null verdict for MercadoLibre is pending the per-firm run. The TOST equivalence test against the ±1.5pp bound (canonical) is reported on the index page's KPI grid; full TOST family on /extensions.

Primary sources

MercadoLibre daily price data

S&P Capital IQ Public Pricing feed — ticker MELI, daily adjusted close, T−260 through T+10 around canonical event date
Cross-check: CRSP via WRDS (academic-grade) — daily adjusted close panel
Factor returns: Ken French Data Library (FF5 + UMD)

Full Bluebook-format citations in SOURCES.md. Replication scripts and expected results in /replication.