Southwest Airlines & the 3% Derivative Threshold

ISS and Glass Lewis have flagged elevated derivative-action thresholds as a governance reduction that disempowers minority shareholders. Their predictable critique of Southwest Airlines' May 16, 2025 bylaw amendment electing into Texas Business Organizations Code §21.552: "Raising the derivative-suit ownership threshold to 3% materially reduces shareholder accountability over the board." Under their framework, the market should price this as a governance-quality reduction — a measurable negative abnormal return on the disclosure date.

The §21.552 election is the first publicly disclosed adoption among publicly traded Texas-incumbent issuers (May 16, 2025 — two days after Texas Senate Bill 29's May 14, 2025 effective date). It is the only adoption to have been judicially tested (Gusinsky v. Reynolds, No. 3:25-cv-01816-K (N.D. Tex. Mar. 17, 2026), dismissed with prejudice). The market response across an exhaustive five-specification battery and a four-airline peer comparison is a precise, measured null. Where ISS and Glass Lewis frameworks predicted a governance discount, the market data show none was priced. That is not the absence of evidence — under two-one-sided-tests (TOST) equivalence framing in Section L.5, it is affirmative evidence that the market did not view §21.552 election as a value-reducing governance change.

Southwest Airlines' Day-0 abnormal return on May 20, 2025 is statistically indistinguishable from zero across every specification we test. Five orthogonal models — market model SPY (+0.092%), JETS-augmented airline-sector control (+0.457%, the methodological headline), SPY+XLI industrials robustness (+0.153%), SPY+USO oil-price robustness (+0.085%), and synthetic control (+0.196%) — all return Day-0 abnormal returns within ±0.5% of zero, with no specification's Patell-z p-value below 0.785. Bottom line: the market priced no measurable §21.552-attributable response on the disclosure date, and the conclusion is robust to whichever factor model a referee or proxy advisor prefers.

Adding the US Global Jets ETF (JETS) raises R² from 0.37 to 0.58 and more than halves the residual standard deviation of the LUV return series (sigma 0.020 → 0.017). For a single-firm event in a highly cyclical sector (airlines), the SPY-only market model materially under-controls for sector-level co-movement. The JETS-augmented specification is the methodologically right headline; the SPY-only model is reported only for cross-cohort comparability. The robustness-footnote rows (SPY+XLI, SPY+USO) are reported only to demonstrate spec invariance — they have weak rationale for a governance event with no contemporaneous industrial-sector or oil-market shock.

The §21.552 election produced no statistically detectable abnormal return on any of three plausible event dates. Tested across (1) the board-adoption Friday, May 16, 2025 — earliest possible information leak (+1.77%, p = 0.39); (2) the actual 8-K filing Monday, May 19 — but after the 4:00 PM market close, so not tradeable (−0.43%, p = 0.84); and (3) the standard next-trading-day convention, Tuesday, May 20 — the headline reading (+0.09%, p = 0.96). Even if a critic chooses any of the three event dates as the "right" one, no reading reaches 10% significance. The board-adoption +1.77% is consistent with random Friday market noise (Friday returns are routinely ±2% in airline equities).

None of the eight cumulative-return windows around the §21.552 disclosure reaches even 10% statistical significance. The closest-to-event reading (Day-0) is +0.09% (p = 0.96); the symmetric three-day window (CAR_−1,+1) is −0.61% (p = 0.86); the broader 21-day window (CAR_−10,+10) is −0.19% (p = 0.98). The largest-magnitude reading — Post +0/+60 days at −12.78% — looks dramatic but is statistically indistinguishable from zero (p = 0.42), reflecting routine noise in airline returns over a three-month horizon. Bottom line: across every window from the immediate disclosure through the +60-day post-event period, no §21.552-attributable signal is detectable in Southwest Airlines' cumulative abnormal returns.

Southwest Airlines' pre-event abnormal-return series shows no statistically detectable trend (slope = +0.06 basis points per day, p = 0.755). This is the placebo check that confirms the Day-0 reading is a clean event response, not the continuation of a pre-existing pattern. If a referee suspects information leakage in the days before the §21.552 8-K filing, this test rules it out: the pre-event period was statistically flat, so any non-zero Day-0 reaction is attributable to the disclosure itself rather than pre-leak drift.

Southwest Airlines slightly trailed the S&P 500 by 4.48 percentage points over the 237 trading days following its §21.552 election (May 20, 2025 through April 29, 2026), but this gap is statistically indistinguishable from zero (Lyon-Barber-Tsai skewness-adjusted p = 0.906). In plain terms: a buy-and-hold investor in Southwest Airlines on the day of the §21.552 disclosure earned +16.51% over the next year, versus +20.99% for the broader market — a 4.48% shortfall that falls well within ordinary year-to-year noise and cannot be attributed to the §21.552 election.

Southwest Airlines OUTPERFORMED 6 of 8 airline peers — including American Airlines, Alaska Air, JetBlue, SkyWest, United Airlines, and Ryanair — over the 237 trading days following the May 20, 2025 §21.552 election. The mean peer-relative buy-and-hold abnormal return was +13.53%; the median was +16.80%. Southwest Airlines underperformed only Delta Air Lines (−14.15%) and Allegiant (−18.69%). No individual pairwise comparison reaches statistical significance — the dispersion across peers is wide (range ≈ 64 percentage points) — but the directional pattern is unambiguous: by the natural-peer benchmark, the §21.552 election was followed by relative outperformance, not punishment.

A single 237-day BHAR pools all events that affect Southwest Airlines' stock price — the §21.552 election, Elliott Investment Management's activist settlement, the "Bags Fly Free" policy reversal, board refresh, fuel-price moves, operational issues, and concurrent litigation. The shortest horizons isolate the §21.552-attributable response because they exclude concurrent events that occurred later. If the shorter-window peer-relative BHARs are also positive (Southwest Airlines outperforming peers), the §21.552-specific reading converges with the longer-window reading — strongest evidence that the bylaw amendment was not punished by the market. If they diverge, the divergence itself tells us which window is being driven by which confound. The two-airline DiD attribution battery in Section K.5.1 estimates the §21.552-clean coefficient directly using firm and date fixed effects.

Southwest Airlines earned an estimated +14.17% annualized abnormal return above what its market exposure would predict over the 12 months following the §21.552 election. The directional sign — positive alpha — is consistent with the peer-cohort BHAR finding that Southwest Airlines outperformed six of eight airline peers (Section E3). The estimate is not statistically distinguishable from zero (t = +0.309, p = 0.757) because 12 monthly observations is well below the ~36 months required for a calendar-time portfolio alpha to be reliably identified at conventional statistical power. This is a "directionally favorable but underpowered" reading — re-run as more post-event data accumulates over the next 24 months.

The pre-fit gate (R² ≥ 0 in levels) fails, meaning the synthetic peer fits LUV's pre-event level path worse than a constant. This is a known failure mode for highly volatile, idiosyncratic stocks (airlines have firm-level cost-structure heterogeneity that makes weighted peer-group counterfactuals hard to construct). The +0.20% day-0 reading is reported for transparency only; under standard SC publication conventions (Abadie 2021), readings from a synthetic that fails the pre-fit gate should not be relied upon for inference. The market-model and JETS-augmented specifications remain the headline-grade evidence.

Once airline-industry effects are properly controlled for, Southwest Airlines' 1-month BHAR after the §21.552 election is essentially zero (+0.38%, p = 0.960). The simpler factor models (CAPM, FF3, FF5, FFC6) all show 4-7% negative BHAR — but that is the airline industry's broad weakness in the post-event window, not a Southwest-specific signal. Adding the airline industry portfolio (JETS) as a seventh factor strips that industry effect out, and the residual collapses to near-zero. The publication-grade FFC6+JETS specification confirms: at the 1-month horizon, no statistically detectable §21.552-attributable abnormal return. Romano-Wolf step-down adjusted p-values are all 1.000 across 20 tests, ruling out any spurious-significance concerns.

Under the publication-grade Carhart-six-factor + airline-industry specification, Southwest Airlines' 12-month buy-and-hold abnormal return is +13.65% (p = 0.602) — directionally positive, not negative. At 1-month +0.38%, at 3-month −13.16%, at 6-month −3.69%, at 12-month +13.65%. None of the four readings reaches statistical significance (all p ≥ 0.31, all RW-adjusted p = 1.000). The 3-month and 6-month negative readings are within the noise band of long-horizon BHAR; the bookend readings (1-month +0.38%, 12-month +13.65%) are positive. For ISS / Glass Lewis review: across the entire 12-month post-event window, the publication-grade specification rules out a §21.552-attributable governance discount and is directionally consistent with the peer-cohort outperformance documented in Section E3.

Against a market-and-airline-industry combined benchmark, Southwest Airlines' 1-month and 12-month BHARs are essentially zero (−0.17% and −2.22%; p = 0.98 and 0.93 respectively). The 3-month and 6-month readings show a transient negative drift (−16.39% and −10.86%) but neither approaches statistical significance. Most importantly, none of the four horizons rules out the null of zero §21.552 effect under any conventional confidence threshold — and the bookend readings (immediate disclosure response and one-year horizon) converge on near-zero.

Under the publication-grade Carhart-six-factor + airline-industry specification, Southwest Airlines' post-event annualized alpha is +18.87% (t = +0.642, p = 0.521). This is directionally positive — Southwest outperformed what its market, size, value, profitability, investment, momentum, and airline-industry exposures would predict. The alternate specifications (FF3, FF5, FFC6 without JETS) all show negative alpha (−20% to −24% annualized), but those negative readings reflect airline-industry weakness loading onto Southwest's stock through its airline-sector beta — once JETS is added as a seventh factor, the residual flips sign to positive. None of the five readings reach statistical significance (all p ≥ 0.50) — consistent with the underpowered nature of a 12-month CTE — but the publication-grade headline is directionally consistent with §21.552 NOT having been priced as a discount.

Southwest Airlines outperformed 7 of 8 airline peers in the immediate post-event month and 6 of 8 over the full 12-month window — including a statistically significant +8.70% outperformance vs. Alaska Air at the 1-month horizon (p = 0.031, the only significant cell in the 32-pair grid). The transient 3-month and 6-month underperformance (mean −16.71% and −1.06% respectively) is driven by the sector-wide airline downturn affecting Southwest's stock through its airline-industry beta — confirmed by the FFC6+JETS specification (Section I.2) where the same horizons show similar negative readings before JETS is added but flip toward zero after. The bookend windows that bracket the §21.552 election (1-month and 12-month) both show Southwest outperforming the peer median by a wide margin, with no statistical evidence of §21.552-attributable underperformance against any single peer.

Single-firm event studies have known identification weaknesses: the day-0 abnormal-return reading depends on the chosen factor model, the chosen estimation window, and the assumption that no firm-specific shock contaminated the event window. Each of the four strategies below isolates a different counterfactual. If the four converge on "no measurable discount," the affirmative-defense claim is robust to the standard identification critiques. If they diverge, the divergence itself is evidence about which threat is operative.

[Pending DiD coefficient] If δ̂ is statistically indistinguishable from zero, Southwest Airlines did not experience a §21.552-attributable discount net of common airline shocks. Under TOST equivalence with δ_eq = 2%, equivalence is rejected if the 95% CI on δ̂ excludes both ±2pp.

[Pending IV result] First-stage F > 10 confirms instrument relevance (TX-incumbent eligibility strongly predicts §21.552 adoption). β̂₁ confidence interval that crosses zero is consistent with no average governance-discount effect across the population of §21.552 elections.

[Pending SDID result] τ̂ that crosses zero in the bootstrap CI confirms no Southwest-specific governance discount net of the synthetic Texas-incumbent counterfactual.

[Pending permutation result] Placebo p > 0.5 means more than half of randomly chosen pre-event dates produce |AR| larger than the real Day-0 reading — i.e., the real reading is unremarkable relative to ordinary trading-day noise.

Each of the four strategies produces a point estimate and a confidence interval. The board's affirmative defense against ISS / Glass Lewis does not require positive evidence of zero effect — it requires evidence that the magnitude of any effect is bounded. A confidence interval that crosses zero AND is bounded within ±2pp under TOST equivalence is the operational standard. The four-strategy convergence test asks: do all four strategies independently produce CI that crosses zero? If yes, no single identification critique is sufficient to overturn the null — the affirmative defense is robust.

Confirmed adopters: 5 of 76 (6.6%) — Tesla, Inc. (TSLA), Southwest Airlines (LUV), CenterPoint Energy (CNP), HeartSciences (HSCS), and Dillard's (DDS) all have Tier-1 (8-K-verified) or Tier-2 confirmation of §21.552 election. CenterPoint's amendment is the most explicit replication of Southwest's bylaw-amendment template — 3% derivative threshold + jury trial waiver + exclusive forum (Texas Business Court). Including Panel-C-v2.5 high-confidence candidates: 6-7 of 76 (7.9-9.2%), depending on how UAMY (reincorporation confirmed but §21.552-election direct-pull pending) and LEGH (no practitioner-press confirmation yet) are scored. The adoption-rate denominator depends on whether one excludes firms that subsequently delisted or were acquired (the active-issuer universe is closer to ~70).

The adoption rate is climbing. One full year after Texas Senate Bill 29 took effect, between 7% and 9% of TX-incumbent public issuers have elected into the new derivative-suit threshold — up from 2.6% as of late April 2026. The post-Gusinsky v. Reynolds period (after March 2026) appears to be when the second wave of adoptions is materializing, consistent with firms waiting for the federal-court guidance before electing. This is itself a publication-grade empirical finding: the adoption pattern is not a single-firm "early mover" phenomenon but a slow-then-accelerating diffusion responding to judicial validation.

Verification protocol: CNP, HSCS, and DDS confirmed via practitioner-press citations + 8-K Item 5.03 reference. LEGH and UAMY remain pending direct-pull EDGAR confirmation. Possible explanations for non-adopters: (a) most TX-incumbent issuers are small enough that a 3% derivative threshold is already effectively unreachable for most plaintiffs at $1M voting-power floor; (b) institutional investors and proxy advisors have signaled opposition to the election, raising shareholder-relations costs; (c) firms are waiting for federal-court guidance on §21.552 enforceability (Gusinsky v. Reynolds, N.D. Tex. Mar. 17, 2026 was the first such test).

The pooled CAAR test asks: across all firms electing into §21.552, did the announcement-day abnormal return on average differ from zero? Two cohort-level findings would be most informative:

• If the Tier 1-2 pool is null (CAAR ≈ 0%, p > 0.10): consistent with LUV's individual finding; suggests the §21.552 election is not material to firm value across the cleanly-identified cases. This would be the strongest available evidence for the null hypothesis.
• If expanding to Tier 3 pulls CAAR away from zero: implies the smaller / less-liquid candidates have larger reactions, which could reflect either (a) genuine heterogeneous treatment effect (smaller firms more affected because derivative-suit risk is more material relative to cap), or (b) confounded readings from small-cap noise around the placeholder event dates.

Pre-registration commitment: the headline cohort statistic is the Tier 1-2 pool with the FFC6+sector specification at the headline T0 for each firm. Tier 3 results are reported as sensitivity. Romano-Wolf step-down adjustment for multiple horizons across the pool-spec matrix is the inferential gold standard.