GNW owns 81% of Enact Holdings (ACT), a best-in-class mortgage insurer. That stake is worth $4.94B. GNW's market cap is $3.4B. The math screams "buy the discount." The factor decomposition says don't.

The Discount

The arithmetic is seductive. Enact stake: $4.94B. GNW market cap: $3.4B. Implied discount: 31%, or -$1.54B assigned to everything outside Enact. The "everything outside Enact" is a $44.1 billion long-term care insurance liability in run-off, a $97M/year CareScout cash burn, $783M in holdco debt (no maturities until 2034), and a $750M litigation windfall the market refuses to price.

The surface-level SOTP looks like a free lunch. It isn't.

The Factor Kill

I ran GNW against its own subsidiary plus market and insurance sector factors:

GNW vs ACT + SPY:       R² = 53.1%  alpha = 0.0%  idio = 46.9%
GNW vs ACT + SPY + KIE: R² = 59.9%  alpha = 0.1%  idio = 40.1%

Alpha is zero across every specification. The 31% discount generates zero excess return. It's not a mispricing converging — it's a structural feature the market prices as permanent LTC negative carry.

More surprising: GNW correlates more with KIE (the insurance sector ETF, r = 0.718) than with ACT (its own 81%-owned subsidiary, r = 0.636). The holdco structure imposes sector beta on top of the Enact exposure. You're buying an insurance sector bet, not a stock pick.

Only 40-47% of GNW's variance is idiosyncratic. The 75% threshold exists because below it, your returns are driven by factors you don't have edge in. GNW fails this badly.

If you want Enact, buy Enact.

The Sinkhole

LTC adjusted operating losses nearly doubled: -$326M (2025) vs -$176M (2024). Liability remeasurement loss: $316M vs $172M. Driver: higher claims, lower terminations. The two largest blocks — Choice I (average age 78) and Choice II (average age 75) — are entering their peak claim years. Peak is 85+. Claims will increase for another 7-10 years as this cohort ages. This is not cyclical. It's actuarial gravity.

Management guided Q4 2025: "averaged $75 million per quarter, continue see losses this level 2026." That's $300M baseline before assumption updates. If assumptions deteriorate again — and they will, as the cohort ages — total LTC losses could exceed $400M in FY2026 (55% probability).

The rate action program is 87% complete ($34.5B of $39.5B NPV), which is genuine progress. But ≈50% of cohorts are capped at 100% net premium ratio — these are the pre-2003 policies generating all the near-term remeasurement losses, and rate actions can't help them further. The other 50% (uncapped, currently profitable) will age into the same peak-claim zone in coming years.

GLIC consolidated RBC: 300%, down from 306%. GLICNY: 207%, up from 200% but still 7 points from the Company Action Level trigger. Management runs legacy subs on a standalone basis — "Not put capital into companies" — but if RBC breaches 200%, regulators may have other ideas.

Cross-ticker check: Unum's closed LTC block is profitable (+$63.5M FY2025), RBC at 425-450%. Prudential's LTC generated $198M. GNW is the worst-positioned LTC carrier in the industry by every metric: deepest losses, lowest RBC, highest proportion of capped cohorts. The claims acceleration is partly industry-wide (all carriers report similar language about "elevated incidence, capped cohorts"), but GNW's severity is idiosyncratic.

The Enact Anchor

Enact is excellent. $558M adjusted operating income (GNW's 81% share). PMIERs at 162% — $1.9B excess capital. FICO 746. Loss ratio 11%. New $500M buyback announced February 2026; GNW gets ≈$405M.

The risk: delinquencies rising (2.62% vs 2.10% two years ago). But this is entirely industry-wide. Essent rose +23bps, Radian +20bps, NMIH +11bps over the same period. Enact's +17bps is actually the smallest increase. Drivers: unemployment at 4.4%, 2020-2022 vintage seasoning, elevated mortgage rates limiting cure. Loss ratios remain historically low vs pre-crisis 50%+ norms. Enact is not broken. It's normalizing.

The structural constraint: Enact EMICO can only pay $3M in "ordinary" dividends in 2026. All capital returns must be structured as share repurchases or extraordinary dividends requiring regulatory approval. This creates execution risk on the $405M expected return.

The Lottery Ticket

AXA/Santander PPI mis-selling litigation. UK High Court ruled July 2025 in AXA's favor: £680M ($911M). Santander is appealing. Hearing: July 21-23, 2026. If upheld, GNW recovers approximately $750M — 22% of market cap.

Management explicitly excluded this from capital allocation plans. The 10-K says recovery would go to "CareScout, share repurchase program, and opportunistically paying down debt." The proceeds flow to the holding company as discontinued operations income — structurally bypassing the insurance subsidiaries. The regulatory trap (regulators forcing proceeds downstream to shore up LTC reserves) is mitigated but not eliminated.

The probability: 62%. UK Court of Appeal overturn rates for commercial cases run 30-40%. High Court findings of fact are rarely disturbed, but damages could be reduced on appeal. Partial success scenarios exist. Full $750M requires complete dismissal. Expected value: $465M, or ≈14% of market cap.

This is 100% idiosyncratic and genuinely under-covered. One active analyst (KBW, $10 target). No social media discussion. No dedicated sell-side research on the litigation outcome. The market appears to price near-zero recovery probability, which is clearly wrong given the High Court already ruled.

But it's a lottery ticket attached to a factor bet. AXA represents maybe 30-40% of GNW's idiosyncratic variance, which itself is only 47% of total variance. So AXA exposure = ≈15-20% of what you're actually buying.

The Dead Signal

Seven insiders "acquired" ≈370K shares ($3.2M) on February 26 — the day before the 10-K filed. Looks like a conviction cluster buy. It isn't. Form 4 transaction codes: M (RSU conversion) + F (tax withholding at $8.62). Annual equity compensation delivery, not open market purchases. Zero bullish signal.

The Buyback Engine

$828M repurchased since May 2022 — 24% of current market cap, ≈6% annual yield. $222M remaining on current authorization, $405M expected from Enact in 2026. Management guided $175M-$225M in buybacks for 2026.

This is the real shareholder story. GNW earns $558M from Enact, loses $414M to Closed Block + Corporate, nets ≈$144M, and returns $175-225M+ via buybacks. They're returning more than they earn, funded by Enact's capital returns. Sustainable as long as Enact stays strong.

The Verdict

The 31% NAV discount is half rational, half fear premium. ≈20% is structural holdco discount (normal for insurance conglomerates). ≈10-15% is LTC uncertainty premium. The exploitable portion depends entirely on two things: whether LTC losses stabilize (trending wrong direction) and whether AXA pays (62% probability, July 2026).

Base case: Enact holds, LTC losses persist at $300-400M/year, buybacks continue, discount stays ≈25-30%. You earn the buyback yield (≈6%) and not much else. Alpha vs financials sector: approximately zero.

Bull case (35%): AXA pays $750M, rate actions complete by 2027, LTC losses peak and begin declining. Discount compresses to ≈20%. Stock reaches $10.50-$14 over 24 months.

Bear case (25%): LTC acceleration overwhelms Enact earnings. RBC breaches trigger regulatory action. Enact capital returns curtailed. Stock to $5.50.

This is a factor bet with an attached litigation lottery ticket. The regression proves it — zero alpha, 40-47% idio, more correlated with the insurance sector than with its own subsidiary. The AXA catalyst is real, under-covered, and the only genuinely idiosyncratic component. But it's 15-20% of what you're buying. The other 80% is insurance sector beta and an LTC sinkhole that's getting worse.