Market evidence · figures 1 through 9

The market didn't blink.

Bottom line. On the day ExxonMobil announced its move to Texas, its stock moved like every other energy company. Three independent statistical methods — matched-peer comparison, synthetic control, and Bayesian posterior — say the same thing: the market priced this change as a non-event. Critics expected a discount for losing shareholder rights. The evidence does not show one.

Figure 1 · Total shareholder return

ExxonMobil versus oil & gas peer composite, last twelve months & since announcement

Tracks peers within ±1 pp across LTM and post-March-10 windows

ExxonMobil's total shareholder return moves with its peer composite both before and after the March 10, 2026 redomiciliation announcement. There is no visible level shift at the event date.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 2 · Peer-firm Day-0 returns

ExxonMobil versus 20 oil & gas peers, March 10, 2026

+0.04 pp matched-pair differential p = 0.958

ExxonMobil's announcement-day return was slightly negative; the 20-peer median was slightly positive. The matched-pair differential is statistically indistinguishable from no effect.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 3 · Co-movement check

ExxonMobil versus Chevron daily returns, t−5 through t+5

Same-sign on 10 of 11 days only t+3 diverges

Across the eleven trading days around the redomiciliation announcement, ExxonMobil and Chevron moved in the same direction on every day except t+3. The two pure-play U.S. integrateds essentially tracked each other through the event window.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 4 · Synthetic-control donor pool

Synthetic ExxonMobil — ADH classical synthetic control, weights sum to 1.000

Top three: Chevron 44.87% · Ovintiv 19.91% · Schlumberger 13.03% ADH unit-simplex (Abadie, Diamond & Hainmueller 2010); 250-day pre-window canonical, 60-day pre-window sensitivity

An SLSQP solver constrained to the unit simplex (donor weights ω ≥ 0, ∑ω = 1) selects a convex combination of the full 20-firm S&P Capital IQ oil & gas universe that minimizes pre-event tracking error to ExxonMobil. Eleven non-zero weights; no firm-level renormalization. The Arkhangelsky Synthetic Difference-in-Differences (SDiD) with time-weighted ridge regularization produces a flatter weight vector (top six: Chevron 9.0%, Schlumberger 8.2%, Halliburton 7.2%, Ovintiv 6.2%, Baker Hughes 5.9%, ConocoPhillips 5.7%). Both estimators are reported in results/donor_weights_canonical.json for full transparency.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 5 · Daily synthetic-control gap

ExxonMobil minus synthetic ExxonMobil, t−20 through t+5

Day-0 gap: +0.15% 95% CI brackets zero across the entire window

The daily gap between ExxonMobil and its synthetic counterfactual hovers around zero throughout the 26-day window. The announcement day produces no visible level break and no excursion outside the pre-event confidence band.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 6 · Cross-firm placebo

Day-0 absolute synthetic-control gaps, 21-firm placebo universe

ADH classical synthetic control, canonical (250-day pre-window): ExxonMobil ranks 6 of 21 · placebo p = 0.286 ADH sensitivity (60-day pre-window): rank 11 of 21, p = 0.500

Replicating the Abadie–Diamond–Hainmueller (ADH 2010) classical synthetic-control procedure on the 21-firm placebo universe (20 donors + treated) produces a distribution of "treatment-day" gaps. Under the 250-day pre-window canonical specification, ExxonMobil's actual gap sits at rank 6 of 21 (placebo p = 0.286). The 60-day pre-window sensitivity check yields rank 11 of 21 (p = 0.500). The Arkhangelsky (2021) time-weighted SDiD variant is reported separately on /extensions (rank 6 of 21 / p = 0.286 canonical; rank 13 of 21 / p = 0.600 sensitivity). All four specifications classify NULL; none places ExxonMobil in the tail of the placebo distribution.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 7 · In-time placebo

Distribution of Day-0 gaps across 100 pseudo-event dates

Actual announcement day: −0.03% near the 50th percentile of placebo dates

Re-running the event study on 100 randomly selected non-event dates in the pre-event window produces a distribution of "treatment-day" gaps. ExxonMobil's actual March 10 gap is unremarkable against that placebo distribution.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 8 · Bayesian posterior

Posterior density over the redomiciliation announcement effect

Posterior mean: +0.02 pp P(effect < −2 pp) = 0.004

A weakly-informative prior centered at zero, updated with the observed market-model abnormal return, yields a posterior tightly centered at zero. The probability that the true announcement effect was worse than −2 percentage points — the threshold for a meaningful expropriation premium — is roughly four in a thousand.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Figure 9 · Three-benchmark comparison

Day-0 abnormal return under three independent benchmarks

Matched pair: raw differential +0.12 pp; market-model adjusted +0.04 pp (t = 0.05, p = 0.958) · XLE −0.03% · SPY+BNO −0.07%

Three benchmark choices — a Chevron matched pair, the XLE energy ETF, and a two-factor SPY+BNO model — produce abnormal-return point estimates clustered within seven basis points of zero. The headline conclusion does not depend on the benchmark choice; multi-specification disclosure is presented in full on the methodology page.

Interactive chart loads when JavaScript is enabled. Headline figures are sourced from the canonical data.json result file.

Three lenses, nine figures, one conclusion

Matched-pair differential

+0.04pp

p = 0.958

Synthetic-control gap

+0.15%

Cross-firm placebo p = 0.500

Bayesian posterior mean

+0.02pp

95% CI [−1.48, +1.53]

P(effect < −2pp)

0.004

Market-discount thesis: no robust support

For the 54-test battery and robustness checks (Patell, Corrado, wild bootstrap, GARCH, HC3, MDE, TOST), see Methodology.

Article of record vs. v1.3 replication kit

Figures on this page reflect the May 17, 2026 v1.3 kit (20-firm donor pool; 250-day pre-window canonical with 60-day pre-window sensitivity; min BH-corrected p = 0.186 across the FF6+BNO four-window family). The May 5, 2026 publication of record (Goodwin, Columbia Law School Blue Sky Blog) reports earlier specifications (10-firm pool, 60-day pre-window, min BH p = 0.82). Both classify NULL. Reconciliation table in reviewer_package/EXTERNAL_RED_TEAM_FINDINGS_2026-05-17.md; article erratum on footnote 27 targeted 2026-06-01.

The formal apparatus

Each of the three analytical lenses (peer comparison, synthetic control, Bayesian) is a textbook estimator with primary-source citations. The formulas below are the operative definitions; the citations are the source authorities. Both reproduce, line-for-line, in the reviewer package.

Lens 1 · Matched-pair peer comparison

Day-0 abnormal return is the difference between ExxonMobil’s actual return and the return predicted by a one-to-one matched peer (Chevron) on the announcement day. Patell prediction-interval inference uses the residual standard deviation from the estimation window adjusted for leverage at the post-event observation.

AR XOM,0 = R XOM,0 - R CVX,0

SE_Patell = σ_ε · √(1 + x′₀(X′X)⁻¹x₀) t = AR₀ / SE_Patell df = n − k − 1

Bluebook citation. James M. Patell, Corporate Forecasts of Earnings Per Share and Stock Price Behavior: Empirical Test, 14 J. Acct. Res. 246 (1976); A. Craig MacKinlay, Event Studies in Economics and Finance, 35 J. Econ. Literature 13 (1997).

Lens 2 · Synthetic-control gap (ADH; Arkhangelsky SDiD)

Donor weights ω minimize pre-period tracking error to the treated firm under simplex constraints (ω ≥ 0, ∑ω = 1). The Day-0 (or post-window-mean) gap is the residual between the treated firm and the weighted donor combination. Arkhangelsky’s SDiD adds time weights λ to absorb pre-trend bias.

ω̂ = argmin_{ω ∈ Ω} ‖ Y_1,pre − ∑_j ω_j Y_j,pre ‖₂² + ζ² T_pre ‖ω‖₂²
τ̂_SDiD = ( &ymacr;_1,post − ∑_j ω̂_j &ymacr;_j,post ) − ∑_t∈pre λ̂_t ( y_1,t − ∑_j ω̂_j y_j,t )

Bluebook citation. Alberto Abadie, Alexis Diamond & Jens Hainmueller, Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program, 105 J. Am. Stat. Ass’n 493 (2010); Dmitry Arkhangelsky, Susan Athey, David A. Hirshberg, Guido W. Imbens & Stefan Wager, Synthetic Difference-in-Differences, 111 Am. Econ. Rev. 4088 (2021); Alberto Abadie, Using Synthetic Controls: Feasibility, Data Requirements, and Methodological Aspects, 59 J. Econ. Literature 391 (2021).

Lens 3 · Bayesian conjugate posterior

Normal–normal conjugate update with a flat prior centered at zero. The pre-period synthetic-control gap distribution (N = 220 trading days) yields the prior variance σ_pre². The Day-0 observed gap updates the posterior; the integrated tail probability quantifies how unlikely an effect more negative than −2pp would be.

Prior: θ ~ N(0, σ_pre²) with σ_pre = 0.77% (from pre-period synthetic-control gap distribution)
Likelihood: AR₀ | θ ~ N(θ, σ_obs²) σ_obs = 0.77% Posterior: θ | AR₀ = +0.021% ~ N(+0.02%, 0.77%²) 95% CI: [−1.48%, +1.53%] P(θ < −2%) = 0.004 P(θ < −1%) = 0.092 P(θ < 0) = 0.486

Bluebook citation. Andrew Gelman, John B. Carlin, Hal S. Stern, David B. Dunson, Aki Vehtari & Donald B. Rubin, Bayesian Data Analysis ch. 2 (3d ed. 2013); Donald A. Berry, Statistics: A Bayesian Perspective 195–221 (1996); Joseph A. Schumpeter Center, Two One-Sided Tests (TOST) for Equivalence, in Schuirmann, A Comparison of the Two One-Sided Tests Procedure and the Power Approach for Assessing the Equivalence of Average Bioavailability, 15 J. Pharmacokin. Biopharm. 657 (1987).

Sources cited on this page

Primary filings. Exxon Mobil Corp., Preliminary Proxy Statement (Form PRE 14A), Accession No. 0001193125-26-098908 (Mar. 10, 2026); Exxon Mobil Corp., Definitive Proxy Statement (Form DEF 14A) at 76–77, Annexes B & C, Accession No. 0001193125-26-147614 (Apr. 8, 2026); Exxon Mobil Corp., Definitive Additional Materials (Form DEFA14A), Accession No. 0001193125-26-219305 (May 12, 2026).
Statutes. N.J. Stat. Ann. §§ 14A:3-6.2, 14A:3-6.3, 14A:3-6.7, 14A:3-6.8, 14A:3-6.9, 14A:5-6, 14A:10A-3(j), 14A:10A-4, 14A:10A-5 (West 2024); Tex. Bus. Orgs. Code Ann. §§ 6.201, 6.202, 21.373, 21.551(2)(C), 21.552(a)(3), 21.553, 21.554, 21.4161, 21.602, 21.606 (West 2025); Del. Code Ann. tit. 8, §§ 203, 228, 327 (2024); S.B. 29, 89th Leg., R.S. (Tex. 2025); S.B. 1057, 89th Leg., R.S. (Tex. 2025).
Event-study methodology. Patell, 14 J. Acct. Res. 246 (1976); MacKinlay, 35 J. Econ. Literature 13 (1997); Eugene F. Fama & Kenneth R. French, A Five-Factor Asset Pricing Model, 116 J. Fin. Econ. 1 (2015); Mark M. Carhart, On Persistence in Mutual Fund Performance, 52 J. Fin. 57 (1997).
Synthetic control and SDiD. Abadie, Diamond & Hainmueller, 105 J. Am. Stat. Ass’n 493 (2010); Arkhangelsky et al., 111 Am. Econ. Rev. 4088 (2021); Abadie, 59 J. Econ. Literature 391 (2021).
Inference under heteroskedasticity and clustering. Russell Davidson & Emmanuel Flachaire, The Wild Bootstrap, Tamed at Last, 146 J. Econometrics 162 (2008); Halbert White, A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity, 48 Econometrica 817 (1980); Tim Bollerslev, Generalized Autoregressive Conditional Heteroskedasticity, 31 J. Econometrics 307 (1986).
Multiple-testing correction and equivalence. Yoav Benjamini & Yosef Hochberg, Controlling the False Discovery Rate, 57 J. Royal Stat. Soc. B 289 (1995); Joseph P. Romano & Michael Wolf, Stepwise Multiple Testing as Formalized Data Snooping, 73 Econometrica 1237 (2005); D.J. Schuirmann, TOST Equivalence, 15 J. Pharmacokin. Biopharm. 657 (1987).
Case law. Gusinsky v. Reynolds, No. 3:25-cv-01816-K, 2026 WL 747179 (N.D. Tex. Mar. 17, 2026) (Kinkeade, J.); TSC Industries, Inc. v. Northway, Inc., 426 U.S. 438 (1976); CSX Corp. v. Children’s Inv. Fund Mgmt. (UK) LLP, 654 F.3d 276 (2d Cir. 2011).