White Mountains Insurance Group, Ltd. WTM

WTM — White Mountains Insurance Group, Ltd. — deep-value report

Generated: 2026-08-28 · Price: … (close 27.08.2026, yfinance) · Market cap: … (2,386,721 shares outstanding at 30.06.2026) / … at stockanalysis.com's current estimated share count (2.36 mn shares) · Sector: Financial Services / P&C Insurance — insurance holding company, domiciled in Bermuda, listed NYSE + BSX · Single currency: USD reporting, USD quotation (no ADR currency risk)

Primary sources: 10-K FY2025 (filed 27.02.2026, CIK 0000776867), 10-Q Q2 2026 (filed 06.08.2026, wtm-20260630.htm — downloaded in this session, was missing from the file), 10-Q Q1 2026 (06.05.2026), 10-Q Q2/Q3 2025, 10-K FY2021–FY2024, DEF 14A of 01.04.2026, 8-K of 26.08.2026. Market data: yfinance + stockanalysis.com. Deterministic data pack: data-pack-WTM-20260828.md. Research brief: research-WTM-20260828.md. USD 10-year sovereign yield: 4.64% (FRED:DGS10, 27.08.2026).

Corrections to the research brief (verified against filings, not taken at face value):

1. The brief gives Q2 2026 net income of "199.5 mn" as "net income attributable to common shareholders" — correct, but consolidated net income is 231.2 mn, of which 31.7 mn goes to minority interests. The distinction matters: ~14… of consolidated profit does NOT belong to the WTM shareholder.
2. The brief reports a consensus of "44 analysts, average target …" (TradingView). False, or more precisely, synthetic. stockanalysis.com shows "n/a" for both analyst count and price target; MarketBeat explicitly shows a consensus target of … and only two ratings dated within the last 15 months (Weiss → Buy, 12.08.2026; Zacks → Hold, 27.05.2026). WTM is effectively uncovered by sell-side. The figure of 44 is an algorithmic aggregate, not a real analyst consensus. Direct consequence for the GBL score — see the final chapter.
3. The brief says "Q4 2025: BVPS … … in Q4." The filing says … for FULL-YEAR 2025, including dividends (10-K FY2025, p. 45). The brief's quarterly figure isn't confirmable from a primary source.
4. The brief doesn't mention the WTM Partners segment as a separately reportable segment. As of Q2 2026, it is — reclassified retroactively (10-Q Q2 2026, p. 57). WTM now has FIVE segments, not four.

---

Executive summary

White Mountains isn't an insurer. It's a capital allocator that owns insurers — a "mini-Alleghany" with 64 employees at the parent holding company (DEF 14A, p. 38: the holding-company bonus pool covers 64 people) controlling USD 13.77 bn of assets. The model, unchanged for decades: buy or build insurance/financial-services platforms at low valuations, leave them fully decentralized under their own management with rollover equity, and when the market pays an absurd multiple, sell them. NSM Insurance sold to Carlyle in 2022 (+~USD 280/share). Bamboo Ide8 sold to CVC in December 2025 for USD 848 mn net cash + 250 mn retained stake, accounting gain 816 mn = ~USD 320/share (10-K FY2025, p. 45). The single declared metric: growth in book value per share. Since the 1985 IPO, 13% annualized in BVPS and 12% in price (DEF 14A, p. 29).

Where it stands today. BVPS at 30.06.2026 = USD 2,257.60 (5,388.3 mn common equity / 2,386,721 shares; 10-Q Q2 2026, p. 56), … in the quarter, … in the half-year, incl. dividends. Price … → P/B = 0.94x. Five-year BVPS CAGR calculated from filings (… at 31.12.2020 → 2,187.97 at 31.12.2025): 11.7%/year; over 3 years (1,457 → 2,187.97): 14.5%/year. The company's own return target, written in the proxy: the 10-year Treasury yield + 700 basis points = 11.64% today. In other words, WTM is delivering exactly its own declared cost of capital — no more, no less.

That's why valuation here is a matter of simple, blunt arithmetic: if normalized ROE ≈ the required rate, intrinsic value ≈ book value, and the entire available discount is the … below book plus any unrecorded franchise value. The residual-income model (Book × [1 + (ROE−r)/(r−g)]) with ROE 11.0%, r 10%, g 3% gives USD 2,580/share (…). The DCF on normalized earnings of USD 430 mn gives USD 2,736 (…). The no-growth EPV gives USD 1,801 (…). The 5-year historical average P/B multiple (0.961x) applied to current BVPS gives USD 2,170 (…). Triangulated range: … … …, median …. Monte Carlo over 20,000 scenarios: median …, P(undervalued) …%, but with a pessimistic tail at … (P10). The gap between the triangulation median (…) and the MC median (…) is NOT a contradiction: the triangulation includes two models that explicitly refuse any growth (EPV) or any re-rating (historical multiples), while the MC is a distribution around a single model, the DCF.

Verdict: HOLD / GRADUAL ACCUMULATION, not a conviction buy. WTM is a genuinely high-quality company, with verifiable evidence of disciplined capital allocation (it buys back stock ONLY below 95% of book — it bought USD 616 mn in 2022 at 0.97x and almost nothing, 7.9 mn, in 2024 when the stock was at 1.11x), with a conservative balance sheet (debt/total capital 13.0%, only USD 8.1 mn due within 12 months) and a book growing over 11%/year. But at 0.94x book, with the property-cat cycle softening at an accelerating pace (rates … to … at 2026 renewals, per Guy Carpenter/Howden), with USD 835 mn of goodwill+intangibles in two untested bets (Distinguished, which is losing 28 mn pre-tax over the half-year, and WTM Partners, i.e. electrical installation contractors), with 58% of shareholders' equity in model-valued assets or a volatile small-cap, and with a CEO of eight months in the role — you don't have the margin of safety that justifies a large position. The 6% discount to book is the correct price of these risks, not an opportunity.

Figure to watch at the Q3 2026 report: Ark's acquisition-cost ratio. It jumped from 23.2% (FY2025) to 27.0% (H1 2026) — the largest margin erosion in the company, larger than the deterioration in the loss ratio, and it is structural (delegated-authority business with higher commission), not cyclical.

---