Methodology · canonical battery adapted for MercadoLibre
Every completed robustness test confirms the headline.
Bottom line. MercadoLibre's event-study battery uses the canonical methodology documented across the SMU CGI Reincorporation Tracker: 250-trading-day estimation window (T−260 to T−11), FF6+BNO factor model as the primary specification, Patell-z and Romano-Wolf step-down for inference, Benjamini-Hochberg FDR for multiple testing, Schuirmann TOST for equivalence, Abadie-Diamond-Hainmueller synthetic control with Arkhangelsky SDiD sensitivity. MercadoLibre's per-firm test inventory and results are pending the canonical run; the framework below is invariant across firms.
Estimation design
Factor models
Three factor models are reported for every firm in the cohort. The primary specification is FF6+BNO; the secondary specifications are FF6 (no oil) and the Chevron matched-pair (oil-sensitivity comparator). For MercadoLibre specifically: BNO inclusion is debatable given MercadoLibre's Latin American e-commerce-not-oil-firm classification, and is reported as a sensitivity rather than primary specification. See /extensions for FF6-only and FF5+UMD variants.
- FF6+BNO (canonical): Fama-French six-factor (Mkt-RF, SMB, HML, RMW, CMA, UMD) plus BNO ETF excess return. Estimation: OLS on the 250-day pre-period.
- FF6 only (sensitivity): same six factors, no oil instrument. Probably the right primary for MercadoLibre given GICS classification.
- Matched pair (consilience): market-model-adjusted Day-0 differential against a single industry peer. For MercadoLibre the analogous comparator is debatable — General Motors, Ford, BYD, Lucid, Rivian — each has limitations. The choice is documented in /extensions.
Inference
- Patell-z (Patell 1976) for parametric Day-0 inference
- Romano-Wolf step-down (Romano & Wolf 2005) for multi-window family-wise control
- Benjamini-Hochberg FDR (BH 1995) for false-discovery-rate control across the 18-cell battery
- Schuirmann TOST (Schuirmann 1987) for affirmative equivalence at ±1.5pp bound
- Wild bootstrap for non-parametric robustness
- GARCH(1,1) (Bollerslev 1986) for leakage-window conditional-variance correction
- HC3 (MacKinnon & White 1985) for heteroskedasticity-robust standard errors
The 54-test inventory
MercadoLibre's per-firm test inventory count: [DATA PENDING]. The canonical battery is 54 tests (XOM template): 18-cell core (3 benchmarks × 6 windows) + 8 stress tests + cohort + placebo + pre-trend + leakage + TOST + robustness. MercadoLibre's count adapts per data availability.
Identification assumptions
- The 250-trading-day pre-period is uncontaminated by anticipation of the redomiciliation announcement (T−1 = January 31, 2024).
- Factor loadings are stable over the pre-period (tested via rolling-window diagnostics on /extensions).
- The marginal trader sets price during the event window. Founder-controller caveat for MercadoLibre: Musk's ~[DATA PENDING — voting %] voting stake may make the marginal-trader assumption non-standard; see /stress_tests.
- The redomiciliation announcement is the only material disclosure on the event date. Tornetta caveat for MercadoLibre: the announcement followed two days after the January 30, 2024 Chancery ruling rescinding Musk's compensation package; the estimation-window pre-period (T−260 to T−11) ends January 16, 2024, before the Chancery ruling; the event-window choice (announcement Feb 1, 2024 vs. effective 2025-06-10) is therefore the load-bearing identification decision and is pre-registered.
Primary sources
Methodology references
- MacKinlay, A. Craig, Event Studies in Economics and Finance, 35 J. Econ. Literature 13 (1997)
- Patell, James M., Corporate Forecasts of Earnings Per Share and Stock Price Behavior: Empirical Tests, 14 J. Acct. Rsch. 246 (1976)
- Romano, Joseph P. & Michael Wolf, Stepwise Multiple Testing as Formalized Data Snooping, 73 Econometrica 1237 (2005)
- Benjamini, Yoav & Yosef Hochberg, Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing, 57 J. Royal Stat. Soc'y B 289 (1995)
- Schuirmann, Donald J., A Comparison of the Two One-Sided Tests Procedure and the Power Approach, 15 J. Pharmacokinetics & Biopharmaceutics 657 (1987)
- Abadie, Alberto, Alexis Diamond & Jens Hainmueller, Synthetic Control Methods for Comparative Case Studies, 105 J. Am. Stat. Ass'n 493 (2010)
- Arkhangelsky, Dmitry, Susan Athey, David A. Hirshberg, Guido W. Imbens & Stefan Wager, Synthetic Difference-in-Differences, 111 Am. Econ. Rev. 4088 (2021)
- Bollerslev, Tim, Generalized Autoregressive Conditional Heteroskedasticity, 31 J. Econometrics 307 (1986)
Full Bluebook-format citations in SOURCES.md.