Selective Q3 2026 Earnings: EPS Up 14% on a 96.8% Combined Ratio as Premiums Shrink 7%

On October 22, 2026, Selective Insurance Group reported third-quarter net income of $128.7 million, or $2.11 per diluted share — up 12% year-over-year — on a 96.8% combined ratio, while net premiums written shrank 7% to $1,122.8 million, after-tax net investment income jumped 16% to $127.3 million, and the board raised the quarterly dividend 12% and authorized a new $200 million buyback. The headline hides the real story: Selective is deliberately writing less business to earn more on what it keeps, and this is the quarter the trade started paying. Shrinking to grow only works if the combined ratio proves it. At 96.8%, it does.

The Headline Numbers​

Every figure below comes from the Form 10-Q and the earnings release.

Revenue fell while earnings rose double digits. That paradox resolves the moment you split the P&L into its two engines: the underwriting engine earned more on less premium, and the investment engine threw off $21.8 million more pre-tax income than a year ago. The net income row is the line the statement prints before the $2.3 million quarterly Series B preferred dividend — our ledger and the figures on our company page use that line in every period, and the row below it shows what remains for common stockholders.

The gap between GAAP EPS ($2.11) and operating EPS ($2.37) is $19.4 million of pre-tax realized and unrealized investment losses flowing through earnings, worth $0.32 a share before a $0.06 tax offset. Strip the marks and operating income grew 32.8% — nearly triple the GAAP growth rate — which is why management led the release with operating ROE of 16.9% against GAAP ROE of 15.1%.

Revenue Deep Dive​

For an insurer, written premiums are the future and earned premiums are the present — today's revenue was substantially sold a year ago. Selective's present is flat; its future is being deliberately pruned, segment by segment:

Standard Commercial is the company — 77% of written premiums — and its 8% decline "driven by targeted retention actions and lower new business" is the quarter's defining decision, not its defining accident. Retention held at 82% while average renewal pure price increased 5.7%, and the release describes the mechanism plainly: "driving higher renewal retention on our best-performing business and meaningfully lower retention on poorer-performing business." That is portfolio surgery with a retention scalpel — keeping the price, firing the customer, on the worst risks. It worked: the segment combined ratio improved 3.2 points to 97.9%.

Standard Personal, 9% of premiums, is the same surgery on a smaller patient: NPW down 8%, renewal pure price up 8.1%, retention 80%, combined ratio down 6.9 points to 103.2%. Still above 100% — the segment loses money underwriting — but the direction and the method match commercial. Excess and Surplus, the profitable 14%, barely shrank (−2% NPW, +1.5% renewal price) and still printed an 86.5% combined ratio, though 10.3 points worse than last year's exceptional 76.2% on what the release calls "normal period-to-period variability associated with property losses."

On the management signal scan, one theme fires loudly and the rest are silent. Sustained pricing momentum is everywhere: "Renewal pure price increases averaged 5.6%" across the book, and CEO John Marchioni's "maintaining underwriting and pricing rigor in a competitive marketplace" plus "executing targeted pricing actions" frame price as the strategy. But there is no demand-exceeds-supply language, no upcycle framing, no robust-demand paragraph — because demand is being turned away on purpose: "remaining disciplined in our approach to new business growth. These actions have affected near-term premium growth, but we believe improving our portfolio's mix and profitability are the foundation for sustainable, profitable growth over time." When an insurer volunteers that growth is suffering for mix, check whether the mix is actually improving. The combined ratios say it is.

The second engine needs no qualification. Net investment income grew 15.7% pre-tax and 16% after tax to $127.3 million, at a 4.4% after-tax yield on both the fixed-income and overall portfolios, generating — the release's striking figure — 15.2 points of annualized ROE in the quarter. With invested assets at $3.54 per dollar of common equity, every quarter of high-for-longer rates is a quarter the float earns like a second underwriting operation.

The Margin Story​

The combined ratio is an insurer's entire P&L in one number — below 100% means the underwriting itself profited. Selective's walk from the disastrous 2024 through today shows a franchise repairing itself in public:

* Computed as (100% − combined ratio) × net earned premiums; the release discloses quarterly underwriting income after tax only ($29.5M vs $13.2M, +124%).

The loss ratio did the work this quarter, improving 3.3 points to 64.6%, and the moving parts underneath split into luck, skill, and arithmetic in roughly equal thirds. Catastrophe losses ran 5.4 points against 2.1 a year ago — roughly $64 million of cats versus $25 million — a 3.3-point headwind. Against it: no prior-year casualty reserve development at the segment level versus 3.3 points (about $40 million) unfavorable last year, non-catastrophe property losses down a point to 13.1, and current-year casualty loss costs down 2.3 points to 46.1. The reserve swing is the single biggest item, and it is genuinely good news — last year's reserves proved light, this year's are holding — but it is also the least repeatable one. The current-year casualty improvement is the number that says the retention surgery is cutting real loss cost.

The expense ratio is the line moving the wrong way: 30.6% to 32.1%, "primarily due to a higher underwriting expense ratio" against the lower loss ratio, as the release puts it. Mechanically, this is the cost of shrinkage — acquisition costs and overhead spread over 7% less written premium and 2% less earned premium. Selective does not break out the pieces, but the direction is structural: if premiums keep shrinking while the company keeps its agents and staff, the ratio rises by arithmetic. Pricing is the offset — renewal pure price up 5.6% is price over volume — but note the cooling trend: renewal pure price ran 9.5% in both FY2024 and FY2025. If pricing keeps converging toward loss-cost trends while volume shrinks, the margin story stalls. That convergence is the number to watch in January.

The One Big Question​

Is the 1.8-point combined improvement underwriting or accounting? Decompose it to the point: −3.3 from the loss ratio (cats +3.3, non-cat property −1.0, prior-year development −3.3, current-year casualty −2.3) against +1.5 from the expense ratio. About half the gross improvement is last year's $40 million adverse reserve charge not repeating — accounting in the sense that it reflects old business, underwriting in the sense that adequate reserves are the product. The other half is current-year loss costs genuinely improving while the company writes less of the worst business. That is the shrinkage thesis working: shed the poorly performing customers, keep the 5.6% price, and the retained book earns a 96.8%.

The GAAP-to-operating bridge answers the companion question — why $2.37 of operating EPS against $2.11 GAAP:

The $19.4 million is marks and realized losses the release strips because, in its words, "the timing of realized investment gains and losses on securities in any given period is largely discretionary." Fair — but the marks cut both ways over time, and last year's $8.1 million of gains flattered the comparison the other direction. Judge Selective on the operating line and the combined ratio together: both moved the right way, and neither needed the bond market's help this quarter.

Tracking a $15.7 Billion Company in Plain Text​

A $15.7 billion balance sheet fits in a text file, and that is the point: double-entry bookkeeping forces every dollar of loss reserves, premium float, and bond holdings to reconcile, so there is nowhere for a bad number to hide. Our ledger follows how we model every company — one zero-sum income transaction per period, balance-sheet assertions against the filing, in millions of US dollars.

Here is the third quarter's income statement exactly as the ledger records it, including the zero-sum check comment — income posts negative (credit), expenses positive (debit), and the offset is the quarter's $128.652 million of net income:

; Check: −1326.300 + 762.696 + 241.356 + 143.828 + 13.215 + 3.117 + 33.436 + 128.652 = 0 ✓
 
2026-09-30 * "Selective Insurance Group, Inc." "FY2026Q3 Income Statement"
  Income:Revenue                         -1326.300 MUSD  ; net premiums earned 1,180.191 + NII earned 160.481 + net realized and unrealized investment losses (19.446) + other income 5.074
  Expenses:CostOfRevenue                    762.696 MUSD  ; Loss and loss expense incurred
  Expenses:PolicyAcquisitionCosts           241.356 MUSD  ; Amortization of deferred policy acquisition costs
  Expenses:SellingGeneralAdministrative     143.828 MUSD  ; Other insurance expenses
  Expenses:Interest                          13.215 MUSD  ; Interest expense
  Expenses:OtherNet                           3.117 MUSD  ; Corporate expenses
  Expenses:IncomeTax                         33.436 MUSD
  Equity:Adjustments                        128.652 MUSD  ; net income offset (RE set by balance assertion); less preferred dividends 2.300 → 126.352 available to common

An insurer's chart needs its own vocabulary, and the ledger carries it: loss reserves sit on InsuranceReserves, unearned premiums on DeferredRevenue (premiums collected, not yet earned), reinsurance recoverables beside prepaid reinsurance premiums, the Series B preferred stock on PreferredStock, and treasury stock as a contra-equity debit swollen by buybacks. A few balance-sheet lines carry the whole quarter:

2026-09-29 pad Liabilities:NonCurrent:InsuranceReserves        Equity:Adjustments
2026-09-30 balance Liabilities:NonCurrent:InsuranceReserves              -8006.495 MUSD  ; Reserve for loss and loss expense
2026-09-29 pad Equity:TreasuryShares                           Equity:Adjustments
2026-09-30 balance Equity:TreasuryShares                                  866.428 MUSD  ; Treasury stock – at cost (contra-equity; $55M / 587,229 shares repurchased in Q3 at $93.66)

The balance-sheet number that tells the whole story is loss reserves: $5,336.9 million at the end of 2023 → $8,006.5 million at September 30, 2026, up 50% in eleven quarters to back an earned-premium base up 41% over FY2023–FY2025. That is float compounding — premiums collected today, claims paid years from now. Reserves plus unearned premiums now total $10.7 billion of float funding an $11.5 billion investment portfolio, and the $9.8 billion available-for-sale bond book inside it is the engine behind the $160.5 million of quarterly investment income. Read alongside treasury stock, up from $635.2 million to $866.4 million over the same span, and the $200 million of fresh buyback authorization, the picture is complete: the float grew, and the owners' slice of it keeps getting bigger.

The Multi-Year Arc​

Three annuals plus the latest nine months show a standard-lines insurer that took its medicine in 2024 and is compounding out the other side:

Two stories run through this table. The first is the 2024 underwriting trough and the climb out: a 103.0% combined ratio and a $132.6 million pre-tax underwriting loss, then 97.2% and $135.9 million of underwriting income a year later, then 96.8% this quarter — a 6.2-point round trip driven by the loss ratio falling from 72.3% to 64.6% while pricing stayed ahead of loss costs. The second is the investment engine doubling underneath: net investment income from $288.2 million in FY2023 to $531.2 million in FY2025 and $453.0 million in nine months, on a bond portfolio that grew with the float and repriced at higher yields. Together they explain why diluted EPS more than doubled from $3.23 to $7.49 last year — and why this year's $5.80 in nine months has it on pace to do it again only if the fourth quarter holds the line.

The Verdict: Bull vs. Bear​

Bull Case

• A combined ratio improving from 103.0% to 97.2% to 96.8% with zero prior-year casualty development this quarter — reserves set in the bad year are holding, and current-year casualty costs fell 2.3 points on the retained book.
• Net investment income nearly doubling in two years: $288.2 million → $457.1 million → $531.2 million → $453.0 million in nine months, at a 4.4% after-tax yield, with full-year guidance just raised to $490 million after tax.
• Pricing power funding discipline: +5.6% average renewal pure price (+8.1% personal) while retention actions shed the worst risks — commercial keeps more of its best business and less of its poorest at an 82% retention rate.
• Capital return at scale: 63% of after-tax net income returned in Q3, a 12% dividend hike marking the thirteenth consecutive annual increase, and $253 million of buyback capacity ($200 million new plus $53 million remaining) against 20.7% debt-to-capitalization.
• E&S prints 86.5% combined — 10 points of margin on 14% of premiums — a profitable growth engine inside a company that is shrinking everywhere else, with earned premiums still growing 1%.

Bear Case

• Net premiums written fell 7% (−8% in commercial, 77% of the book); earned premiums already slipped 2%, and two more quarters like this turns discipline into shrinkage while the expense ratio loses its denominator.
• The underwriting expense ratio rose 1.5 points to 32.1% on the smaller premium base — the arithmetic cost of shrinkage — and Selective discloses no plan to cut overhead to match.
• About half the year-over-year combined improvement is last year's $40 million adverse reserve charge not repeating; the other supposed tailwind, cats, actually worsened 3.3 points, and full-year guidance assumes only 5.5 points of cats with hurricane season in the rearview but winter ahead.
• Renewal pure price momentum is cooling fast: +5.6% now versus +9.5% in both FY2024 and FY2025 — if it converges toward loss-cost trends while volume shrinks, both levers of the margin story stall at once.
• Rates cut both ways: accumulated other comprehensive loss of $491.8 million is 14% of equity, book value fell 5% in the quarter on unrealized bond losses, and personal lines still underwrite at a 103.2% combined ratio — the turnaround's weakest link.

Our Take

Buy the discipline, watch the denominator. Selective just posted the quarter its strategy promises for a competitive market: shedding underpriced premium on purpose, holding price on what it keeps, earning a 96.8% combined ratio on the retained book, and converting the margin into a double-digit dividend hike and buybacks. The reserve swing flatters the comparison but does not invalidate it — current-year loss costs improved while cats worsened, which is underwriting, not accounting. The risk that matters is written premiums: the expense ratio is already charging for the shrinkage, and pricing momentum is cooling from 9.5% toward 5.6%. If Q4 stabilizes premiums near flat while holding 96–97%, this is a 16%-operating-ROE compounder with $253 million of buyback dry powder. If NPW prints −7% again, the thesis becomes a bet on cost cuts management has not named. Until then, the shrinking is working.