CINF — Cincinnati Financial Corporation

Deep-value analysis (REFRESH): Cincinnati Financial Corporation (CINF)

Generated: 2026-09-04 · Reference price: … (yfinance/tracker, 09/04/2026) · Market: NASDAQ Global Select · Sector: P&C + Life Insurance · Market cap … · Shares outstanding 153.45mn (06/30/2026)

ticker: CINF
regime: REFRESH
baseline_thesis: 2026-08-02
verdict: SPECULATIVE
prior_verdict: SPECULATIVE
confidence: high
falsifier_triggered: pe (DEFINITION ARTIFACT - see Executive Summary)

New primary sources versus the 08/02/2026 analysis: Q2 2026 10-Q (filed 07/27/2026, accession 0000020286-26-000045), the Q2 2026 earnings release (8-K 0000020286-26-000043, exhibit 99.1, 07/27/2026), the dividend announcement from 08/21/2026 (8-K 0000020286-26-000047), the investor handout from 08/25/2026 (8-K 0000020286-26-000049 — 17 slides, images only, no extractable text), the CINF data pack 09/04/2026, Monte Carlo simulation mc-CINF-20260904.json, analyst consensus (stockanalysis.com / marketscreener, Sept. 2026).

Inherited sources, not reread (unchanged by the new filing): 10-K FY2021–FY2025, 10-Q Q1–Q3 2025, DEF 14A from 03/18/2026, AM Best rating March 2026. The “The business and the moat,” “Management and capital allocation” and “CEO profile” chapters from the 08/02/2026 report are structurally inherited and updated at specific points where Q2 2026 moved something (buybacks, portfolio rotation, per-agency production); the full SEC file wasn’t reread.

Executive summary

What triggered the resumption and why the trigger is false. The mechanical deep_delta.py triage flagged the pe falsifier as BROKEN: the thesis said “the stock is overvalued on normalized operating earnings,” with an invalidation threshold of pe < 15, and yfinance now reports a P/E of 8.05–8.25. The figure is real, but it measures something different from the thesis. yfinance’s P/E uses trailing-12-month GAAP EPS, and at CINF GAAP EPS fully contains the mark-to-market swing of a … common-stock portfolio. Over the Q3 2025 – Q2 2026 window, GAAP net income was … (~…/share; the tracker shows 21.55), of which approximately … was after-tax investment gains. Without them, non-GAAP operating profit — exactly the metric the company publishes and analysts model — was …, i.e. …/share. The relevant P/E isn’t 8.1×, it’s 17.9× on realized TTM and 20.9× on the normalized base derived below. The pe falsifier cannot test this thesis, because at CINF the number it reads swings between 8× and 25× based entirely on how the U.S. stock market did in the reference quarter. The falsifier that DOES track the thesis — pb < 1.2 — holds: P/B is 1.61 (… / … BVPS). debt_equity also holds (total debt/capital ratio 4.6%, down from 4.9%), as does fcf_yield (…%, but see below why that’s a meaningless figure at an insurer).

What actually changed over the five weeks. A single new quarter, Q2 2026, but one that changes the thesis’s composition without changing its conclusion. Non-GAAP operating profit fell 28% YoY, to …mn (…/share) from …mn (…), on a combined ratio of 100.8% versus 94.9% — the first quarterly underwriting loss (…mn) after a long streak. What kept the quarter in positive territory is investment income: … pretax YoY, … on bond interest alone. In other words, the thesis has shifted from “expensive relative to underwriting” to “expensive relative to underwriting, but with a genuinely accelerating investment engine.” Both components are new in magnitude; the overall direction isn’t.

The owner-earnings base, stated explicitly, because it’s the only thing that matters in the valuation. CFO − capex — the … and the …% yield the data pack shows — is NOT owner earnings at an insurer. In H1 2026 alone, CFO of … contains the increase in loss reserves (+…mn) and unearned premiums (+…mn): money collected from policyholders for future claims, i.e. float, not shareholder money. The correct base is non-GAAP operating profit, normalized for catastrophe load. Starting from TTM …, …mn is subtracted (normalizing catastrophes from the realized 7.5 pp to the five-year average of 9.0 pp) and …mn (resetting H2 2025 — a half-year with a current-accident-year ex-catastrophe combined ratio of 84.6% — to the current pace of 87.8%) → …, i.e. …/share. Mandatory external check: analyst consensus for 2026 EPS is … — my base is within 0.4% of the market, not off by an order of magnitude, so the AFYA-type error isn’t repeated.

Estimated value. The five-model triangulation gives a margin-of-safety range from … (EPV Greenwald, no-growth annuity) to … (BVPS × total ROE over a full cycle), with the median at …. The Monte Carlo simulation (20,000 scenarios on the same DCF model as the tracker, with my assumptions) gives a median intrinsic value of … and a median MOS of ** …, with only ** … probability the stock is undervalued and …% probability of a margin above 30%.

Verdict: SPECULATIVE — unchanged, but with a less extreme thesis than in August. The triangulation base moved from … to …, and the normalized operating multiple from 22.3× to 20.9×, not because the price fell (… in 32 days), but because operating earnings genuinely grew: after-tax investment income rose from …mn (FY2025) to … (TTM) and is accelerating. It remains a structurally high-quality company — a Dividend King with 66 consecutive years of increases, debt/capital 4.6%, A+ Superior at AM Best — bought at a price that leaves no classic Graham/Buffett margin of safety on recurring earning power. What deteriorated versus August: commercial underwriting (combined ratio 104.1% in Q2, +11.2 pp YoY), favorable reserve development (1.7 pp in Q2 2026 versus 2.6 pp in Q2 2025), new business (…) and, most importantly for the Thorndike grid, buybacks accelerated 9.4× right at all-time price highs (…mn in H1 2026 versus …mn in H1 2025, at an implied ~1.7–1.8× book value).