The Allstate ALL

EARNINGS BANNER — Q2 2026 reports 2026-08-05, ONE DAY away

Allstate reports second-quarter results tomorrow, 2026-08-05, on consensus EPS of $6.06 and revenue of $15,458 million. The last five prints beat consensus EPS by 40.1%, 82.8%, 45.6%, 45.6% and 45.7% — an extraordinary and, as Section 3 shows, largely explainable run. Christian M. Lown started as Executive Vice President and Chief Financial Officer on 2026-08-03, two days before the release.

The verdict below is set with that timing as an explicit input. The stock closed at $262.59 with RSI at 72.8, 10.8% above its 50-day average, 4.3% below its 52-week high, and — unusually — with a published street price-target consensus of $258.46 and median of $261, BOTH BELOW the current price. A company that has beaten by 40%+ five times running, into a print tomorrow, with the sell side's own targets already beneath the price, is a company where the risk is asymmetric against the holder in the short run, and this dive does not take a new position into it.

Reference table

1. What the business is

The Allstate Corporation, headquartered in Illinois with 53,300 employees, run by chairman and chief executive Thomas Joseph Wilson, provides property, casualty and other insurance across the United States and Canada. The profile block describes four segments: Allstate Protection, Protection Services, Allstate Health and Benefits and Run-off Property-Liability — "standard private passenger auto and homeowners policies, as well as specialized vehicle coverage for motorcycles, trailers, motor homes, and off-road vehicles... renter, condominium" cover and more.

That description is now partly out of date, and the 10-Q says why:

> "The dispositions of the employer voluntary benefits ('EVB') and group health businesses in 2025 did not qualify for discontinued operations. The Allstate Health and Benefits segment is no longer a reportable segment, with results of this segment recast to reflect only the results of the EVB and group health businesses. The retained individual health business, previously included in the Allstate Health and Benefits segment, is a non-reportable segment with results included in all other for all periods presented."

This is a real corporate action with a visible fingerprint in the payload and it is not identified there. seg_prod shows Allstate Health And Benefits revenue collapsing from $2,538 million in FY2024 to $676 million in FY2025 — a 73% fall that no field explains. cf_a shows acquisitionsNet of POSITIVE $3,129 million in FY2025 against negative figures in every other year — that is divestiture proceeds, not acquisition spend, and a reader treating it as acquisitions would conclude Allstate was a large net buyer of businesses in a year it was a large net seller. And the Q1 2026 income statement confirms the effect: accident and health insurance premiums and contract charges of $136 million against $487 million a year earlier.

Revenue by segment, from seg_prod ($M):

seg_prod reconciles to within 3.8% in FY2025 and 0.9% in FY2024, which is acceptable but not exact, and the FY2025 gap is the recast. Read the table and Allstate is a personal-lines property-casualty insurer with a fee-services attachment. Property-Liability at 89.9% determines everything. Protection Services — vehicle service contracts, roadside assistance, identity and consumer protection — grew 28% over two years to $3,550 million and is the only part of the business with fee rather than underwriting economics.

Geography. seg_geo stops at FY2023 and shows United States $54,688M, Non-United States $2,060M and European Union $346M. It correctly identifies the United States as the largest market by a factor of 26 — a check that failed on one name in the previous batch — but there is no data covering FY2024 or FY2025, and the block is not used for anything current.

2. The March quarter, and what actually produced it

Consolidated statement of operations, three months ended 31 March, $M, filing-verified:

The whole quarter is in one line: claims and claims expense fell $1,630 million while premiums rose $855 million. Everything else is secondary.

The Property-Liability underwriting table — the most important table in this dive

Three months ended 31 March, $M and ratios, filing-verified:

Now the arithmetic that decides this dive, stated as plainly as it can be.

The combined ratio improved 15.4 points. The filing attributes 7.3 of those points to a lighter catastrophe load and 5.0 to larger prior-year reserve releases. That is 12.3 of 15.4 points — 80% of the improvement — from two items that are not a run rate.

On catastrophes. Allstate published $1,240 million of catastrophe losses in the quarter against $2,202 million a year earlier. The company issues a monthly press release specifically to disclose estimated catastrophe losses — two of the three 8-Ks in this archive, dated 2026-06-18 and 2026-07-16, exist for no other purpose. A business that reports a number monthly because it is too volatile to leave to quarters is not a business whose good quarter should be annualised.

On reserve development. A prior-year reserve reestimate is a revision to an earlier estimate of claims already incurred. A $1,004 million release means Allstate concluded it had over-reserved past accident years by that amount, and the entire sum flows to current-period income. It is real money and it is genuine evidence of conservative past reserving. It is not this quarter's underwriting, and a release of that size — 6.8 points of combined ratio, four times the prior year's — cannot repeat indefinitely, because reserves are finite.

The normalised picture, labelled as our arithmetic and not the company's. Restore the prior year's 15.7-point catastrophe load and the prior year's 1.8-point reserve benefit and the Q1 2026 combined ratio would be approximately 94.3. On $14,802 million of premiums earned, the difference between 82.0 and 94.3 is roughly $1,820 million of underwriting income — taking the quarter from $2,658 million to approximately $838 million. That is not a bear case; it is the same quarter with average weather and average reserve development.

We do not claim the true mid-cycle number is 94.3. Catastrophe loads vary, Allstate has been raising homeowners rates, and the expense ratio at 21.3 is genuinely good. What we claim is narrower and harder to argue with: the headline 82.0 contains roughly 12 points that a shareholder should not capitalise, and the consensus estimate row proves the sell side agrees.

3. The beat record, and why it is less mysterious than it looks

Five consecutive beats between 40% and 83%. Three of them at almost exactly 45.6%.

A run like this is either a company systematically outperforming or an estimate process that cannot model the variable driving results, and here it is plainly the second. Analysts model a normal catastrophe load and normal reserve development; Allstate has delivered five quarters of below-normal catastrophes and above-normal releases. When the driver reverses, the same estimate process will produce misses of similar magnitude in the other direction. That is the honest reading of a 45.6% beat repeated three times, and it is why this dive treats the beat record as a warning rather than as evidence of quality.

Note also the reporting-basis gap. The Q1 2026 epsActual of $10.65 in earn_cal is an adjusted figure; the 10-Q's GAAP diluted EPS is $9.25, a 15.1% difference driven mainly by $405 million of net losses on investments and derivatives. All GAAP figures in this dive come from the 10-Q; the earn_cal figures are used only for the surprise comparison, on the adjusted basis the estimates are struck on.

4. Balance sheet and capital

From the 10-Q at 2026-03-31 and bal_a at 2025-12-31:

The balance sheet is genuinely strong and the growth in book value is the most durable good news in this file. $113.52 of book value per diluted share, up 52.2% year on year and 4.7% in a single quarter, against total debt of $7,490 million and equity of $31.61 billion. Interest coverage of 39.4x and a dividend payout ratio of 9.6% mean the company retains almost everything it earns.

Two cautions on reading that book-value growth. First, a 52.2% increase in book value is arithmetically the same peak-cycle phenomenon as the 48.4% return on equity, and it will decelerate with the combined ratio. Second, at $262.59 the stock trades at 2.31x that book value — a full multiple for a personal-lines insurer, which historically supports 1.5x to 2.5x depending on where the cycle sits. At the top of a cycle, 2.31x book is not a discount.

Capital return. Weighted-average diluted shares fell from 268.8 million to 262.6 million year on year, a 2.3% reduction; cf_a reports $1,233 million of FY2025 repurchases against $2 million in FY2024. The buyback restarted in 2025 after being effectively suspended. The dividend is $4.16 trailing, a 1.58% yield on a 9.6% payout ratio. Preferred dividends run at $29 million a quarter, $116 million a year, on $2,001 million of preferred stock the vendor's enterprise-value calculation ignores.

Cash-flow measures are not used as a lens here. The vendor reports FY2025 operating cash flow of $10,110 million, free cash flow of $9,882 million and a freeCashFlowYieldTTM of 17.06%. For an insurer, operating cash flow includes the growth of the float — premiums collected against claims not yet paid — and bears no relation to distributable earnings. freeCashFlowYieldTTM, priceToFreeCashFlowRatioTTM (5.86x), evToFreeCashFlowTTM (6.45x) and incomeQualityTTM (1.11) are all rejected as lenses for this name. The relevant measures are the combined ratio (82.0), return on average common equity (48.4%), book value per diluted share ($113.52) and the catastrophe and reserve-development components of the combined ratio — all filing-sourced.

5. Valuation — priced in or room?

At $262.59 (market cap $67.60B, approximately 257.4M shares):

The estimate row is the single most informative thing on this page and it needs stating without softening: consensus expects Allstate's earnings per share to FALL from $31.34 in FY2026 to $26.358 in FY2027, a decline of 15.9%, on thirteen analysts in each year. The trailing multiple of 5.6x is therefore not a valuation; it is a peak-earnings artefact. The forward multiples of 8.4x and 10.0x are the honest ones, and 10.0x on a declining earnings estimate is a fair rather than a cheap price.

Estimate dispersion is wide and informative. The FY2027 EPS range is $23.809 to $31.596 — a 32.7% spread, one of the widest in this batch — which is exactly what you would expect when the variable in question is catastrophe frequency. FY2028 rests on 3 analysts and is used only as a cross-check.

est.ebitdaAvg and est.ebitAvg are REJECTED, on two independent grounds. First, the fabrication signature: from FY2024 through FY2028, ebitdaAvg is exactly 9.581% of revenueAvg and ebitAvg exactly 8.232%, in every year — five consecutive years at constant margins. Second, netIncomeAvg EXCEEDS ebitAvg in every forward year — FY2026 shows net income of $8,253 million against EBIT of $5,108 million, and FY2027 $6,781 million against $5,343 million — which is arithmetically impossible. Both defect signatures documented in the data contract fire on this file simultaneously. All forward valuation uses epsAvg.

Enterprise value — the preferred stock is omitted. enterpriseValueTTM of $74,390M less the market capitalisation of $67,596M implies $6,794M, against netDebt of $6,812M at 2025-12-31. The $2,001 million of preferred stock is not added. A corrected figure is approximately $76,391M, 2.7% higher. For an insurer, enterprise value is a weak concept anyway — $85.16 billion of investments backing policyholder liabilities is neither debt nor cash in the usual sense — and no EV-based multiple is used in this dive. The vendor's own investedCapitalTTM of MINUS $48,972 million producing a returnOnInvestedCapitalTTM of +31.2% demonstrates that its capital framework does not accommodate this balance sheet.

Peer context. The vendor peer set is, for once, correctly constructed: Aflac, Chubb, Cincinnati Financial, Hartford, Progressive, Travelers and W. R. Berkley — seven genuine insurance comparables including the two closest, Progressive and Travelers. No peer multiple comparison is drawn because this file holds no verified current estimates for those names, but the set itself is sound and is recorded as such. Note that two of the three names in the knowledge base's only claim — Travelers and Berkshire — sit in or adjacent to this set.

5a. What today's price assumes (the inversion)

At $262.59 — 8.4x FY2026 consensus, 10.0x FY2027, 2.31x filed book value — the price embeds:

• EPS of $31.34 in 2026 and $26.36 in 2027. (Consensus; 13 analysts each.) The first quarter delivered $10.65 on the adjusted basis, so the year needs $20.69 across the remaining three quarters against a Q2 consensus of $6.06. Note the price already embeds a 15.9% decline into 2027; the question is not whether earnings fall but by how much.
• The combined ratio settles somewhere in the high eighties rather than at 82.0 or at 97.4. (Our derivation from the filed components.) This is the most fragile assumption in the price. At the prior year's catastrophe load and reserve development the Q1 figure is approximately 94.3; at the reported 82.0 the company earns roughly $2.7 billion a quarter from underwriting alone. The gap between those two states is worth more than $7 billion of annual pre-tax income, and no forecast in this file resolves it.
• Catastrophe losses stay meaningfully below the 2025 run rate. (Our number, from the filing.) Catastrophes cost 8.4 points of combined ratio in Q1 2026 and 15.7 points in Q1 2025. Each point is roughly $148 million of quarterly pre-tax income on current premiums earned.
• Prior-year reserve releases continue at some fraction of $1,004 million a quarter. (Our number.) At 6.8 points of combined ratio, this is the largest single discretionary contributor to reported earnings, and reserve redundancy is a finite stock. If releases go to zero, the combined ratio rises 6.8 points with no change in the underlying business.
• The market keeps paying roughly 2.2-2.4x book and 10x forward earnings. (Our number.) At 1.9x the filed book value the stock is $216; at 2.6x, $295. Book-value multiples for personal-lines insurers compress hard when the cycle turns, and the 2.31x currently paid is at the upper end of the historical band.

5b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (MINUS 15.9%, FY2026E $31.34 to FY2027E $26.36) + multiple expansion (8.4x on the forward year to about 11.0x on the then-forward year, +31%) + dividend yield (+1.58%) ≈ +10% to +12%.

That decomposition should be read twice, because it is the opposite of the shape this programme prefers. The base case requires the market to pay a materially HIGHER multiple on materially LOWER earnings. That is not an unreasonable expectation — it is precisely what happens when a cyclical passes its peak and the market begins capitalising mid-cycle rather than peak earnings — but a thesis whose entire positive return comes from multiple expansion against a declining earnings line is a weak thesis, and naming it as such is the point of this section.

The book-value cross-check produces the same answer by a different route, which is mildly reassuring. Book value per diluted share is $113.52 and grew 4.7% in the March quarter. At a more normal retention rate — FY2027 consensus EPS of $26.36 less a $4.16 dividend, less buybacks — book value might reach roughly $130-135 by the end of 2027. At 2.2x that is $286-297, which brackets our $290 base. Two independent methods landing within 2% of each other is the strongest thing that can be said for this valuation.

If the multiple stayed at 8.4x FY2027E the price would be $221 (−15.8%). At 12x, $316 (+20.4%).

5c. Variant perception (where we differ, what would surprise)

• We are ABOVE the street on price and BELOW it on quality, which is an unusual combination and we say so. Our base of $290 is 12.2% ABOVE the consensus target of $258.46, which itself sits 1.6% BELOW the current price, with a median of $261 also below spot. We think the street's target is too low because it is anchoring too tightly to peak earnings falling, and we think the street's rating distribution — 24 holds out of 44 — is exactly right. A price target below the current price alongside a Hold consensus is the sell side saying "good company, done for now," and we agree with the conclusion while disagreeing modestly with the level.
• We think the reserve release is under-discussed relative to the catastrophe story. Everyone can see that catastrophes were light. Fewer people are looking at the $1,004 million prior-year reserve release, four times the prior year's $251 million, worth 6.8 points of combined ratio against 1.8. Watchable number: prior-year reserve reestimates in the 2026-08-05 release. If they are again above $750 million the earnings quality question sharpens rather than resolves; if they fall toward $250 million the combined ratio deteriorates by roughly 5 points with nothing having changed in the business.
• We think the beat record is being read backwards by the market and we read it as a warning. Five consecutive beats of 40-83%, three of them at almost precisely 45.6%, is a signature of an estimate process that cannot model catastrophe frequency, not of a company systematically outperforming. The same process will miss by similar magnitudes when weather normalises.
• We hold no differentiated view on the underlying business, which is well run. Premiums written +2.3%, premiums earned +5.5%, an expense ratio of 21.3 that barely moved, book value compounding, a 9.6% payout ratio, a restarted buyback and a 0.237 debt-to-equity ratio. On the operating company we are at consensus and we say so.
• Positive surprise that would force a re-rate: a Q2 combined ratio below 88 with catastrophe losses at or above 12 points and reserve releases below 3 points — that is, a good number achieved on a normal weather and reserve backdrop, which would be genuine evidence that the underlying loss ratio has structurally improved; or a large buyback authorisation.
• Negative surprise that would break the thesis: a Q2 combined ratio above 95; catastrophe losses above $2.5 billion in the quarter; prior-year reserve reestimates turning to a strengthening rather than a release; the FY2027 consensus falling below $24; or a break of the 50-day average at $237.01 on company-specific news. Additionally, and specifically: any accounting or disclosure change in the first quarter reported under a chief financial officer who started on 2026-08-03.

Synthos fair values

All three anchors are multiples of the FY2027 consensus EPS distribution (mean $26.358, low $23.809, high $31.596, 13 analysts), each cross-checked against the filed book value per diluted common share of $113.52.

• Bear ~$200 — 8.4x the FY2027 consensus LOW of $23.809, and 1.76x current book value. Cross-check: 6.4x FY2026E; 5.3% above the 52-week low of $190.00. The scenario: catastrophe losses return to or above the 2025 load, reserve releases stop, the combined ratio moves back above 95, the FY2027 consensus falls toward $22, and the market applies a trough book multiple to trough earnings. −23.8%.
• Base ~$290 — 11.0x the FY2027 consensus MEAN of $26.358, and 2.55x current book value — approximately 2.2x an estimated end-2027 book value of $130-135. Cross-check: 9.3x FY2026E. Sensitivity, stated openly: 10x FY2027E gives $264 — spot — and 12x gives $316. The whole answer lives in a two-turn band on an estimate row whose own analyst range spans 32.7%. The scenario: the combined ratio settles in the high eighties, catastrophes are ordinary, reserve releases moderate rather than stop, book value compounds and the market pays a mid-cycle multiple. +10.4%.
• Bull ~$345 — 10.9x the FY2027 consensus HIGH of $31.596, and 3.04x current book value. Cross-check: 25.7% above the 52-week high of $274.39 and 8.2% above the highest published street target of $319. The scenario: the loss-ratio improvement proves structural rather than meteorological, homeowners rate increases keep earning through, Protection Services scales, and the market re-rates Allstate toward the multiple its returns on equity would justify if they were durable. +31.4%.

Base is 10.4% above spot; asymmetry roughly 1.32:1 (23.8% down, 31.4% up), plus a 1.58% dividend. A base case near 10% with a payoff ratio below 1.5:1, on a cyclical at the top of its cycle, one day before a print, with the street's own target below the current price and RSI at 72.8, is the definition of Hold.

6. Knowledge base — 1,955 raw hits, one survivor, and the largest discard in this programme

Raw hits: 1,955. After a case-sensitive entity re-run: 1. Name-level claims on Allstate: 1, and it is BULLISH. Discarded: 1,954.

This is the largest raw sweep this programme has ever produced and the least useful, by a wide margin in both directions. The sweep included the bare token ALL alongside Allstate and The Allstate Corporation, plus free text on property and casualty and auto insurance, across all 52,021 distilled claims. It returned 1,955 hits across 91 channels — led by jordi_visser (180), raoul_pal (101), darius_dale (87), arthur_hayes (67), forward_guidance (64), jensen_huang (64), real_vision (64) and luke_gromen (58).

Essentially every one of those 1,954 discards is a claim containing the ordinary English word "all."

This is the homograph defect in its purest available form, and it is worth pausing on because the numbers make the general point better than any previous instance. The programme has documented CEG matching Constellation Software on 26 of 29, EMR matching only Emerson Collective, USB matching Universal Serial Bus, and MMM matching "$3M" strings. Those produced dozens of wrong hits. This produced 1,954. A sweep with a 99.95% false-positive rate does not degrade gracefully — it produces a large, confident, entirely fabricated "consensus" on any name a writer cares to construct one for. The entire sweep is discarded, and the discard is the finding.

The single surviving claim, verbatim:

> 2026-07-28 · BULLISH · conviction 70 · horizon: thesis · entities: BRK.B, TRV, ALL · channel: Josh Brown · speaker: Josh Brown · role: independent

> "Been pounding the table on insurance (Travelers, All State) making record highs; Berkshire is firing on all cylinders — Coca-Cola, Amex, transports, utilities — it has everything investors want now."

Three things about it, in order of usefulness.

It carries a NAMED SPEAKER, and the channel is the speaker. Most claims in this store carry speaker: null and are attributed only to a channel — the weakest sourcing the 4-lane policy admits, and the failure mode that produced the 2026-07-25 attribution crisis. This one is attributed to Josh Brown as both channel and speaker with role: independent, and the content confirms it: this is independent commentary, not management voice. That is the strongest attribution available in this batch.

It is seven days old — the freshest name-level claim across all twelve names in this batch. Dated 2026-07-28 against a dive struck 2026-08-04.

And it is a momentum observation, not a valuation argument. "Making record highs" is a description of the chart. The claim names three entities — Berkshire, Travelers and Allstate — as an insurance-and-quality basket that "has everything investors want now," which is a statement about flows and positioning. It says nothing about the combined ratio, nothing about the reserve releases, nothing about the catastrophe load and nothing about the price. It is a real, current, well-attributed positive and it does not engage with a single one of the four things this dive concluded on. Weighted accordingly: positive-very-low.

Conclusion. Breadth 1, claim count 1, net conviction positive-very-low. The Synthos knowledge base holds one claim on Allstate, it is bullish, it is well sourced and recent, and it is a momentum call. No concentration sensitivity test is meaningful on a single claim from a single speaker.

7. Data integrity — what we rejected and why

Six findings, plus a documented corporate action the payload records only as an unexplained anomaly.

1. The 2025 divestiture of the employer voluntary benefits and group health businesses is recorded only as unexplained data movements. The 10-Q states the dispositions "did not qualify for discontinued operations" and that "the Allstate Health and Benefits segment is no longer a reportable segment." In the payload this appears as: seg_prod showing Allstate Health And Benefits revenue falling from $2,538M to $676M with no explanation; and cf_a reporting acquisitionsNet of POSITIVE $3,129M for FY2025 against −$814M, −$180M and −$288M in the three prior years. A positive acquisitions line is divestiture proceeds, not acquisition spend, and a reader treating it as spend would conclude Allstate was a large net buyer in a year it was a large net seller. This is the "corporate action absent from the payload" class, present in the diagnostic-fingerprint variant.

2. est.ebitdaAvg and est.ebitAvg trip TWO independent documented defect signatures simultaneously — REJECTED. (a) Fixed-ratio fabrication: from FY2024 through FY2028, ebitdaAvg is exactly 9.581% of revenueAvg and ebitAvg exactly 8.232%, in every year. (b) netIncomeAvg exceeds ebitAvg in every forward year — FY2026 net income of $8,253M against EBIT of $5,108M, FY2027 $6,781M against $5,343M, FY2028 $6,416M against $5,428M — arithmetically impossible, and the signature the data contract measures at 8.9% of files. Additionally the FY2022 and FY2023 rows carry NEGATIVE ebitAvg (−$574M and −$845M) alongside positive ebitdaAvg, which is possible but which combined with the above disqualifies the block. All forward valuation uses epsAvg. The FY2028 row rests on 3 analysts and is used only as a cross-check.

3. bookValuePerShareTTM does not deduct preferred equity — CONFIRMED, and it matters here. The vendor reports $121.769, and shareholdersEquityPerShareTTM of $121.847 is essentially identical, confirming no preferred deduction is applied. The 10-Q's own figure is $113.52 of book value per DILUTED common share at 2026-03-31, after deducting the $2,001 million of preferred stock and using diluted rather than basic shares. The vendor overstates by 7.3%, taking price-to-book from a true 2.31x to a reported 2.16x. This is the same defect confirmed on COF in an earlier batch. The filing's $113.52 is used throughout.

4. Preferred stock of $2,001 million is OMITTED from enterprise value. enterpriseValueTTM of $74,390M less market capitalisation of $67,596M implies $6,794M, consistent with netDebt of $6,812M and with the noncontrolling interest of −$21M. The $2,001M of preferred is simply not added. A corrected figure is approximately $76,391M, 2.7% higher. No EV-based multiple is used in this dive in any case, because for an insurer with $85.16 billion of investments backing policyholder liabilities the concept does not apply — as the vendor's own investedCapitalTTM of MINUS $48,972 million producing a positive 31.2% return on invested capital demonstrates.

5. bal_a reclassifies large amounts between line items across years, making the three-year series non-comparable. Between FY2024 and FY2025, intangibleAssets moves from $754M to $6,721M while goodwill FALLS from $3,245M to $3,118M; longTermInvestments moves from $4,718M to $78,350M; totalCurrentLiabilities moves from $0 to $70,936M; and shortTermDebt from $0 to $550M. The 10-Q reports total investments of $83.24 billion at 2025-12-31, so the $78,350M "long-term investments" figure is a reclassification of the investment portfolio rather than a new asset. The FY2023 and FY2024 rows report zero current liabilities for a company with $90 billion of total liabilities, which is not a description of reality. The balance-sheet series is used only for total debt, preferred stock and equity, each cross-checked against the 10-Q where possible.

6. seg_geo stops at FY2023 — no data covers FY2024 or FY2025. The most recent row shows United States $54,688M, Non-United States $2,060M and European Union $346M. It correctly identifies the United States as by far the largest market, which is a check that failed elsewhere in this programme, but it is two years stale and is not used for any current figure.

Verified CLEAN — recorded because clean checks are findings:

• capitalExpenditure appears CLEAN and free cash flow is computed in the right DIRECTION. FY2025 capex of −$228M against D&A of $482M gives capexToDepreciationTTM of 0.184. That is low in absolute terms but entirely plausible for an insurer with almost no fixed assets — fixedAssetTurnoverTTM of 110.8 confirms a tiny property base. Critically, freeCashFlowOperatingCashFlowRatioTTM reads 0.985 — correctly BELOW 1.0, so capex was subtracted, not added, and operatingCashFlowPerShareTTM $45.135 less capexPerShareTTM $0.678 equals freeCashFlowPerShareTTM $44.456 to three decimals. The arithmetic is right. The resulting 17.06% free-cash-flow yield is nonetheless REJECTED as a lens, because for an insurer operating cash flow includes growth in the float and bears no relation to distributable earnings — the same treatment applied to banks in this programme.
• seg_prod reconciles acceptably, summing to $63,959M in FY2025 against consolidated revenue of $66,460M (3.8% gap, explained by the recast) and $64,104M in FY2024 against $63,515M (0.9%).
• The peer set is CORRECTLY CONSTRUCTED — Aflac, Chubb, Cincinnati Financial, Hartford, Progressive, Travelers and W. R. Berkley, seven genuine insurance comparables including the two closest. Against peer sets in this programme that have contained payments processors, oil majors and a company name attached to the wrong ticker, this is worth recording.
• tech.max_dd_from_peak of −4.300% coincides exactly with pct_from_hi. Per the data contract this is correct, not a defect — the six-year peak falls inside the last twelve months. Described as the current distance from the high, never as a maximum drawdown.
• quote versus tech. tech.last of $262.59 equals quote.price of $262.59 exactly. quote.yearHigh/yearLow ($277.22/$188.08) against tech.hi52/lo52 ($274.39/$190.00) — a 1.0% discrepancy on both. tech is used throughout for internal consistency.
• dividendPerShareTTM of $4.16 matches profile.lastDividend of 4.16 and is consistent with the 9.6% payout ratio on trailing EPS of $46.82. Clean.

Vendor composite rating — A / 4 overall, the highest in this batch, with returnOnEquityScore and returnOnAssetsScore both 5 out of 5. Those two sub-scores are computed on a 48.4% return on equity that the company's own filing shows to be a cyclical peak inflated by low catastrophes and reserve releases. The rating block is accurate about the trailing data and is not a forward statement; it carries no weight in this verdict. The priceToBookScore of 2 is the one sub-score pointing at the concern.

Non-equity tripwire — checked and passed. ALL is common stock, NYSE-listed. Price of $262.59 is not par-like; beta is 0.166; volume was 1.07M shares (~$281M of turnover); the 52-week band of $190.00 to $274.39 is a 44% range; the dividend is a regular quarterly common dividend. Note the separate preferred stock — $2,001 million carrying $29 million of quarterly dividends — which is NOT this security. This is common equity.

8. Technicals

• Price $262.59. −4.3% from the 52-week high of $274.39; +38.2% above the 52-week low of $190.00. Position within the annual range: 86th percentile.
• Both moving averages are well below the price and the shorter is above the longer. +10.8% above a 50-day average of $237.01; +22.4% above a 200-day average of $214.52. A clean, established uptrend.
• RSI 72.8 — OVERBOUGHT, the only reading above 70 in this batch. MACD +7.88 — strongly positive.
• Current distance from the six-year peak: −4.3%, identical to the 252-day figure. Per the data contract that coincidence is correct.
• Relative performance: 3-month +19.4% against SPY +7.6% and QQQ +7.7% — outperforming by nearly 12 points; 6-month +30.8% against SPY +11.1%; 12-month +31.4% against SPY +24.3% and QQQ +30.8%. Allstate has beaten the index over every window in the file, which for a 0.166-beta insurer is unusual and is entirely explained by the earnings.
• Sentiment, and it is the most striking sentiment reading in this batch: 1 strong buy, 18 buy, 24 hold, 1 sell across 44 analysts, with the vendor consensus label reading HOLD. Consensus target $258.46 — 1.6% BELOW spot. Median $261 — also below spot. High $319 (+21.5%), low $213 (−18.9%). A consensus target beneath the current price is uncommon and is the sell side stating plainly that the move is complete. Elsewhere in this programme we have treated "every target above spot" as a caution; the honest symmetric treatment of "consensus target below spot" is to treat it as a genuine negative signal, and we do.

Today's move and what it does to the entry

ALL closed 2026-08-04 at $262.59, down $0.09 or 0.03%, from $262.68. It opened at $257.26, traded $255.11 to $263.17, and closed near the day's high on 1.07M shares — light volume ahead of tomorrow's report. No filing is dated 2026-08-04; the most recent is the 8-K of 2026-07-16 carrying the June catastrophe-loss release.

The honest read: this is a fully-priced stock in a strong uptrend, one day before a binary event. RSI 72.8, 10.8% above the 50-day, 4.3% below the 52-week high, with the sell side's own price target already beneath the price. There is no entry argument here and no urgency to exit either — which is precisely what Hold means.

9. Insiders — eight filings, and a new chief financial officer who is not in any of them

There is not one open-market purchase and not one open-market sale in the file. Three director awards priced at $243.12, a director restricted-stock-unit conversion, and one officer vesting with a 29.4% tax withholding at $221.01. All mechanical. No signal.

Two things are worth extracting.

The reference prices document the move. Director awards were priced at $243.12 on 2026-07-01 and the officer withholding at $221.01 on 2026-06-05. The stock closed at $262.59 on 2026-08-04 — 8.0% above the July price and 18.8% above the June one.

And the most consequential personnel event of the last month appears in no Form 4 at all. From the 8-K filed 2026-07-14: "effective August 3, 2026, Christian M. Lown will join Allstate as Executive Vice President and Chief Financial Officer... Prior to joining Allstate, Mr. Lown, age 56, served as the chief financial officer of CoStar Group, Inc. from July 2024 to July 2026 and executive vice president and chief financial officer of the Federal Home Loan Mortgage Corporation (Freddie Mac) from June 2020 to June 2024." His package: $875,000 base salary, a cash incentive target of 200% of salary, an equity opportunity of 375% of salary (60% performance shares, 20% restricted units, 20% options), a $2,000,000 cash sign-on bonus and a $4,100,000 sign-on equity grant vesting ratably over three years.

He started on 2026-08-03, one day before this dive and two days before the second-quarter earnings release. A chief financial officer with two days in the job presenting a quarter he did not oversee is not a red flag, and it is a real reason to expect the 2026-08-05 communication to be more cautious than the numbers might otherwise support. Note also that the prior chief financial officer's departure is not documented in this archive; the 8-K announces an arrival, not a replacement, and the filings available here do not say what happened.

10. Verdict, kill-criteria and flip conditions

Hold.

The case for the company, at full strength. Allstate has produced the best operating results in its history. Combined ratio 82.0 against 97.4. Underwriting income $2,658 million against $360 million. Net income to common $2,428 million against $566 million. Return on average common shareholders' equity of 48.4% over twelve months, against 21.4%. Book value per diluted common share of $113.52, up 52.2% year on year and 4.7% in a quarter. Five consecutive earnings beats of between 40% and 83%. Total debt of $7,490 million against $31.61 billion of equity, interest coverage of 39.4x, a 9.6% payout ratio, a restarted buyback that took the diluted share count from 268.8 million to 262.6 million, and a beta of 0.166. The knowledge base's only claim is bullish, seven days old and from a named independent speaker. This is a well-run company and nothing in this document disputes that.

The case against the price, which is what decides the verdict, in four parts.

First, and decisively, the filing itself disaggregates the quarter. Of the 15.4 points of combined-ratio improvement, 7.3 came from a lighter catastrophe load (8.4 points against 15.7) and 5.0 from larger prior-year reserve releases ($1,004 million against $251 million, worth 6.8 points against 1.8). That is roughly 80% of the improvement from two items nobody controls and neither of which is a run rate. Normalise both and the combined ratio is nearer 94.3, and the quarter's underwriting income nearer $838 million than $2,658 million.

Second, the consensus already says so. EPS of $31.34 for FY2026 falling to $26.36 for FY2027 — a modelled 15.9% decline — on thirteen analysts in each year. The trailing 5.6x multiple is a peak-earnings artefact; the honest figure is 10.0x FY2027, which is fair and not cheap.

Third, the sell side's own price target is below the price. Consensus $258.46 and median $261 against a close of $262.59, with 24 of 44 ratings at Hold. We have consistently treated unanimity about the absence of downside as a caution elsewhere in this programme; the symmetric and honest treatment of a consensus target beneath spot is to treat it as a real negative, and we do.

Fourth, the timing is against a new position. RSI 72.8, 10.8% above the 50-day, 4.3% from the high, up 19.4% in three months — one day before a print, two days after a new chief financial officer started.

Our base of $290 is 10.4% above spot on 1.32:1 asymmetry. That is a fair-value marker, not an opportunity.

Pre-registered KILL criteria — what would take this to Avoid:

• A Q2 2026 combined ratio above 95, or above 92 on a normal catastrophe load, indicating the 82.0 was entirely meteorological.
• Prior-year reserve reestimates turning to a STRENGTHENING rather than a release, which would remove 6.8 points of benefit and add a charge on top.
• Catastrophe losses above $2.5 billion in any quarter, against $1,240 million in Q1 2026 and $2,202 million in Q1 2025.
• The FY2027 consensus EPS falling below $24, against $26.358 today with a low of $23.809.
• Book value per diluted share declining in any quarter, against $113.52 and rising.
• A break below the 50-day average of $237.01 on company-specific news, which on a stock 22.4% above its 200-day average would leave a long way to the next support.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A Q2 combined ratio below 88 achieved WITH catastrophe losses at or above 12 points and reserve releases below 3 points. This is the single cleanest trigger in this document: it would demonstrate that the underlying loss ratio has structurally improved rather than that the weather was kind.
• The FY2027 consensus EPS being revised UP above $29, which would mean the street has concluded the margin improvement is durable.
• A price below approximately $225, which is 8.5x FY2027E and 1.98x current book value, at which the base case exceeds 28%.
• A large share-repurchase authorisation, at a 9.6% payout ratio with 48.4% returns on equity and $31.61 billion of equity.
• Protection Services disclosed at above $4.5 billion of revenue with segment margins, which would begin to change the multiple the whole company deserves.
• A second independent knowledge-base claim engaging with the underwriting rather than the chart. The one that exists is a momentum observation.

Where ALL fits in the Synthos Framework Portfolio. No position today, with an explicit calendar trigger: revisit within three days of the 2026-08-05 print, and specifically on the catastrophe and reserve-development lines rather than on the headline. The financials sleeve would take this name at a target weight of 2% on either of the first two upgrade conditions or at a price near $225. Sizing note: this is a cyclical at the top of a cycle with an excellent balance sheet, which means the correct discipline is a price trigger rather than a story trigger. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $262.59, with the fair-value anchors, the kill criteria and the upgrade conditions all gradeable.

Single biggest risk: normalisation of the two items that made the quarter. Catastrophe losses contributed 8.4 points to the March combined ratio against 15.7 a year earlier. Prior-year reserve reestimates released $1,004 million, worth 6.8 points against 1.8. Together those two lines account for roughly 12 of the 15.4 points of improvement, and each point of combined ratio is approximately $148 million of quarterly pre-tax income on $14,802 million of premiums earned. Restore the prior year's levels and the quarter's underwriting income falls by roughly $1.8 billion. The company publishes a catastrophe-loss estimate every single month, which is the clearest possible statement by management that this variable is too volatile to be left to quarterly reporting — and two of the three 8-Ks in this archive exist for that purpose alone. The consensus estimate row has already priced a 15.9% earnings decline into FY2027 and the sell side's price target already sits below the market price. Allstate is a very good insurer having a very good year, and at 10.0x a falling forward estimate and 2.31x a book value inflated by the same cycle, the price is a fair one rather than an opportunity.

Provenance & disclosures

• Traceability: 1 name-level knowledge-base claim names Allstate out of 52,021 distilled claims (raw hits 1,955 — the largest sweep this programme has produced, surviving a case-sensitive entity re-run 1, entity matches 1, discarded 1,954; breadth 1, net conviction positive-very-low). The claim: Josh Brown, 2026-07-28, bullish, conviction 70, horizon thesis, speaker Josh Brown, role independent, entities ['BRK.B','TRV','ALL'] — "Been pounding the table on insurance (Travelers, All State) making record highs; Berkshire is firing on all cylinders — Coca-Cola, Amex, transports, utilities — it has everything investors want now." It carries a NAMED SPEAKER, which most claims in this store do not and which makes it the best-attributed claim in this batch; it is seven days old, the freshest name-level claim across all twelve names; and it is explicitly a momentum observation — "making record highs" — from a commentator naming Allstate as one of three constituents in an insurance-and-quality basket. It engages with none of the four items this dive concluded on: the combined ratio, the catastrophe load, the reserve releases or the price. The 1,954 discards are the purest instance of the documented homograph defect this programme has recorded — a sweep including the bare token "ALL" matched claims containing the ordinary English word "all" across 91 channels led by jordi_visser (180), raoul_pal (101), darius_dale (87), arthur_hayes (67), forward_guidance (64), jensen_huang (64), real_vision (64) and luke_gromen (58). A 99.95% false-positive rate does not degrade gracefully; it manufactures a confident fabricated consensus for any writer who does not run the case-sensitive re-run. The entire sweep is discarded and the discard is itself the finding. No concentration test is meaningful on a single claim. The quote is verbatim from the stored claim text.
• Data as-of: fundamentals — income statement, underwriting table, combined ratio, catastrophe losses, prior-year reserve reestimates, shareholders' equity, book value per diluted share and return on equity through 2026-03-31, all filing-verified from the 10-Q filed 2026-04-29; annual figures for FY2025, FY2024 and FY2023 from the vendor payload, with the FY2025 segment recast confirmed against the 10-Q · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873598 = 2026-08-04T19:59:58Z ($262.59, −0.03%; 50-DMA $237.01; 200-DMA $214.52; RSI 72.8; MACD +7.88) · knowledge-base claims 2026-08-04. Allstate's fiscal year is the calendar year. There is no second-quarter 10-Q — results are released 2026-08-05, one day after this dive — so the latest filed financials are at 2026-03-31. All figures come from the Synthos vendor data file for ALL or from the SEC filings in the ALL archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-20 (fiscal 2025); 10-Q filed 2025-11-05 (September 2025 quarter); 10-Q filed 2026-04-29 (March 2026 quarter — source of the Property-Liability underwriting table, the combined-ratio components, the catastrophe and prior-year-reserve disclosures, the balance-sheet highlights, the book-value-per-diluted-share figure and the segment-recast language); 8-K filed 2026-06-18 (Item 7.01 — the May 2026 monthly release of estimated catastrophe losses and policies in force, furnished as Exhibit 99, the exhibit content itself is not in this archive); 8-K filed 2026-07-14 (Item 5.02 — the appointment of Christian M. Lown as Executive Vice President and Chief Financial Officer effective 2026-08-03, with full compensation terms); 8-K filed 2026-07-16 (Item 7.01 — the June 2026 monthly catastrophe-loss release, exhibit content likewise not in this archive). All carry preserved [TABLE] statement data — 120 tables in the March 10-Q.
• Where the filings contradicted or corrected the vendor (detailed in Section 7): the 2025 divestiture of the employer voluntary benefits and group health businesses, which the 10-Q states "did not qualify for discontinued operations" and which caused the Allstate Health and Benefits segment to cease being reportable — visible in the payload only as seg_prod revenue falling from $2,538M to $676M with no explanation and as acquisitionsNet of POSITIVE $3,129M for FY2025, which is divestiture proceeds and not acquisition spend; bookValuePerShareTTM of $121.769, which does not deduct the $2,001M of preferred stock and which the 10-Q's own $113.52 per diluted common share exceeds by 7.3%, taking true price-to-book from a reported 2.16x to 2.31x; enterpriseValueTTM, which omits the same $2,001M of preferred, understating by 2.7%; est.ebitdaAvg/ebitAvg, which trip TWO documented signatures simultaneously — fixed at exactly 9.581% and 8.232% of revenueAvg in every year from FY2024 to FY2028, AND with netIncomeAvg EXCEEDING ebitAvg in every forward year, which is impossible; bal_a, which reclassifies large amounts between line items across years — intangibleAssets from $754M to $6,721M, longTermInvestments from $4,718M to $78,350M, totalCurrentLiabilities from $0 to $70,936M — making the three-year series non-comparable; and seg_geo, which stops at FY2023. Where vendor and filing AGREED — recorded, because clean checks are findings: capitalExpenditure is clean in direction, with freeCashFlowOperatingCashFlowRatioTTM of 0.985 correctly BELOW 1.0 and operatingCashFlowPerShareTTM less capexPerShareTTM equalling freeCashFlowPerShareTTM to three decimals — though the resulting 17.06% yield is rejected as a LENS, not as arithmetic, because for an insurer operating cash flow includes growth in the float; seg_prod reconciles to within 3.8% in FY2025 and 0.9% in FY2024; the peer set is correctly constructed (Aflac, Chubb, Cincinnati Financial, Hartford, Progressive, Travelers, W. R. Berkley); dividendPerShareTTM of $4.16 matches profile.lastDividend and the 9.6% payout ratio; tech.last equals quote.price exactly and the 52-week extremes agree to within 1.0%; and tech.max_dd_from_peak of −4.300% coincides with pct_from_hi correctly.
• Basis note: the earn_cal EPS figures are on an adjusted basis and are used in this dive only for the surprise comparison. The Q1 2026 epsActual of $10.65 corresponds to a GAAP diluted EPS of $9.25 in the 10-Q, a 15.1% difference driven mainly by $405 million of net losses on investments and derivatives. All GAAP income-statement figures come from the filing. Forward multiples use consensus epsAvg, which is on the adjusted basis. The trailing multiple of 5.6x is computed on the vendor's trailing per-share figure of $46.82 and is a peak-earnings artefact rather than a valuation, for the reasons set out in Sections 2 and 5.
• Insurance-metric note: cash-flow-based measures (freeCashFlowYieldTTM 17.06%, priceToFreeCashFlowRatioTTM 5.86x, evToFreeCashFlowTTM 6.45x, incomeQualityTTM 1.11) are not used anywhere in this dive, because for an insurer operating cash flow includes the growth of the float and bears no relation to distributable earnings. Enterprise-value measures are likewise not used, as the vendor's own investedCapitalTTM of MINUS $48,972 million producing a positive 31.2% return on invested capital demonstrates. The measures used are the combined ratio (82.0), the loss ratio (60.7), the expense ratio (21.3), the catastrophe-loss effect on the combined ratio (8.4 points), the prior-year reserve-reestimate effect (6.8 points), underwriting income ($2,658M), return on average common shareholders' equity (48.4%) and book value per diluted common share ($113.52) — all filing-sourced.
• Estimate coverage: 13 analysts on FY2026 EPS; 13 on FY2027 — the anchor for all three fair values, with a range of $23.809 to $31.596, a 32.7% spread that is among the widest in this batch and is exactly what one would expect when the swing variable is catastrophe frequency; 3 on FY2028, used only as a cross-check. Revenue estimates rest on only 5 analysts in every forward year.
• Peer note: the vendor peer set — Aflac, Chubb, Cincinnati Financial, Hartford, Progressive, Travelers, W. R. Berkley — is correctly constructed for this name, containing seven genuine insurance comparables including the two closest, Progressive and Travelers. No peer-multiple comparison is drawn because this file holds no verified current estimates for those names.
• Fair-value caveat: the $200 / $290 / $345 anchors are multiples of the FY2027 consensus EPS distribution — 8.4x the low of $23.809, 11.0x the mean of $26.358, and 10.9x the high of $31.596 — each cross-checked against the filing's book value per diluted common share of $113.52 (1.76x, 2.55x and 3.04x respectively). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 10x FY2027E gives $264, which is spot; 12x gives $316. Section 5b states explicitly that the base case requires the market to pay a HIGHER multiple on LOWER earnings, and names that as the thesis's weakest leg. A second, independent route — 2.2x an estimated end-2027 book value of $130-135 — gives $286-297, bracketing the $290 base, and two methods agreeing within 2% is the strongest support available for this number. The base of $290 is 12.2% ABOVE the street consensus of $258.46, which itself sits 1.6% BELOW the current price.
• Timing — a verdict input, not a footnote: second-quarter 2026 results are released 2026-08-05, ONE DAY after this dive, on consensus EPS of $6.06 and revenue of $15,458M. The last five prints beat consensus EPS by 40.1%, 82.8%, 45.6%, 45.6% and 45.7% respectively — a pattern this dive reads as an estimate process unable to model catastrophe frequency rather than as systematic outperformance. Christian M. Lown started as Executive Vice President and Chief Financial Officer on 2026-08-03, one day before this dive and two days before the release; the departure of his predecessor is not documented in this archive. The most recent insider filings are dated 2026-07-07 and earlier and contain no open-market transaction of any kind. 2026-08-04 carried no company filing.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.