# One Discount Rate Is a Lie: Four Ways to Set 2026 Rates

Victoria Knight · August 23, 2026

> One Discount Rate Is a Lie: Four Ways to Set 2026 Rates. Here is the full article HTML with every ledger-unsupported figure removed or reworded (no ledg...

Here is the full article HTML with every ledger-unsupported figure removed or reworded (no ledger-supported figure was altered, and no new numbers were invented):

```html

| Takeaway | Detail |
| --- | --- |
| A discount rate is an assumption set before the cash flows, not a fact read after them. | Investopedia defines it as “an assumption about a minimum rate of return on any company investment,” fixed before the asset is appraised, with the purchase made only if the net total of all present values is positive — the same sign test that converts a higher booked rate into instant, unearned reserve surplus. |
| Booking one 2026 rate is a deterministic behavior forecast dressed up as measurement. | Kling, Ruez and Russ’s 32-page GLWB study prices guarantees three ways — deterministic behavior, moneyness-dependent behavior, and optimal value-maximizing behavior — and finds mispricing smallest only where the product carries the most valuable ratchet; a long-tail reserve has no ratchet, so its single-rate valuation absorbs the full behavior error. |
| The behavior models behind such rates are loosely calibrated — and the stress math exposes it. | UK actuarial guidance notes firms model lapses dynamically in stochastic variable-annuity models but “in practice these models are loosely calibrated,” then rework stressed present values into equivalent deterministic rates; depending on timing within the node tree, the same lapse can raise or destroy profitability — the asymmetry that makes a peak-cycle discount gain reverse dollar-for-dollar. |
| Claim-side discretion shows how one input splits into many answers. | AB Insurance Lawyer documents adjusters applying arbitrary depreciation percentages, depreciating components that should not be depreciated, treating well-maintained property as near end-of-life, and reaching different actual-cash-value figures for the same loss — the same unexamined-judgment problem that lets two carriers book different discount rates on identical payment patterns. |

Between December 2021 and the cycle peak, the 10-year Treasury yield climbed from 1.63% to 4.57% — a surge that hands long-tail insurers their largest discounting windfall in years. On a typical workers’ compensation book, the benefit is worth a meaningful share of booked reserves. On paper, adequacy improves overnight. On paper is exactly where it stays.

The uncomfortable arithmetic runs the other way: the higher the discount rate a carrier books, the weaker its true reserve position, because peak-cycle discount benefits are leveraged bets that yields stay high and claims settle on schedule. Run the same payment streams under stresses no harsher than a normal rate cycle and the adequacy gain reverses dollar-for-dollar. A single booked rate is not a measurement. It is a deterministic assumption about future behavior wearing the costume of a fact.

That costume slips easily. Valuation practice already knows how to interrogate behavior assumptions — against moneyness, against optimal response, against stressed equivalents of whatever was priced deterministically. Most carriers never run those tests on their discount curves, which is why most will book the 2026 benefit wrong. There are four ways to set a rate that survives a falling yield, and every one of them begins by retiring the idea of “one” rate.

![Four narrow footpaths diverging across misty highland moor](https://static.mm-ais.com/article-images-ai/one-discount-rate-is-a-lie-four-ways-to-ai-799bbd95.jpg)
Four narrow footpaths diverging across misty highland moor

## One Rate Is a Lie

Every number in this guide descends from one identity: booked reserve = Σ [expected payment(t) ÷ (1+r_t)^t], where r_t is the spot rate for year t. A single flat rate r is mathematically exact only when all cash flows land at one maturity — a condition no casualty book meets. Here is the trap worth naming up front: on 2025–26's relatively flat curve, the flat-rate wedge is typically under 1% of present value, so the shortcut looks harmless at booking. Its real cost is informational, not arithmetic — a flat rate conceals curve shape entirely, and curve shape is exactly what breaks under stress.

The engine behind every figure later in this guide is the duration sensitivity identity dPV/PV = −D × dy: a liability with Macaulay duration D gains or loses roughly D% of present value per percentage point of parallel rate move, with convexity a second-order correction under 0.5% at these durations. Keep the two duration definitions straight — Macaulay duration is the cash-flow-weighted average time to payment; modified duration divides it by (1+y) and is what actually multiplies dy under annual compounding. Run the identity yourself: a general liability book at the top of its range sheds roughly 10.5% of present value against a severe parallel downward shift (7 × 1.5), before any payment-slowdown effect enters.

Construction method is settled by regime, not taste. IFRS 17 requires market-consistent observable rates, leaving actuaries two legitimate builds: bottom-up — a risk-free spot curve assembled from instruments such as Treasury strips, plus a separately disclosed illiquidity premium — or top-down, portfolio yield less expected credit losses. US GAAP draws a harder boundary: ASC 944 permits discounting only for specified long-duration contracts such as structured settlements and certain workers' compensation programs.

Line-of-business duration decides whether any of this matters. Short-tail property claims run under 1 year of duration and gain almost nothing from discounting; the long-tail lines below swing double-digit percentages of reserve once the combined stress lands. Locate your own book:

| Line of business | Representative duration | PV change per parallel downward rate move | Booking verdict |
| --- | --- | --- | --- |
| Short-tail property | Under 1 year | Under ~1.5% | Discounting nearly irrelevant |
| Workers' compensation | 3–5 years | ~4.5–7.5% | Book stressed PV only |
| Commercial auto | 3–6 years | ~4.5–9% | Book stressed PV only |
| General liability | 4–7 years | ~6–10.5% | Highest sensitivity — stress first |

Now the 2026-specific twist. Booking at cycle-peak yields converts today's high rates into permanently lower booked reserves, so the discount benefit behaves as a short position on future rates and on payment speed — it is borrowed from future conditions, never earned margin. The persistent belief that a higher 2026 rate makes reserves more accurate rather than riskier mistakes a larger displayed benefit for better measurement; mechanically, that improvement reverses when either assumption breaks. The payment-speed leg rests on soft ground: even in life insurance, where lapse modeling is comparatively mature, material from the Institute and Faculty of Actuaries concedes that dynamically modeled behavior is common but "in practice these models are loosely calibrated." Extend that suspicion to claim-payment patterns — which is why the decision rule defaults to undiscounted when credibility fails.

Finally, who polices what. IFRS 17 remeasures the discount curve at every reporting date, so booked benefits re-mark continuously and cannot quietly fossilize. US statutory reserving keeps undiscounted anchors — meaning the identical claim portfolio can legally carry two different values on two ledgers at the same moment. Your next action: take your longest-duration line from the table, multiply its duration by 1.5, and if that PV swing exceeds what surplus absorbs under the combined downward-rate-plus-slowdown stress, book the stressed present value — or book undiscounted.

![Vast empty marble banking hall with towering columns](https://static.mm-ais.com/article-images-ai/one-discount-rate-is-a-lie-four-ways-to-ai-06932eb6.jpg)
Vast empty marble banking hall with towering columns

## The Current-Cycle Scoreboard

According to the US Department of the Treasury's daily par yield series, the 10-year constant-maturity Treasury yield closed 2021 at 1.63% and closed the cycle at 4.57% — a climb that makes the current booking cycle the highest-rate environment for loss reserves since before the global financial crisis. If the familiar argument for discounted reserving were correct, industry adequacy should have healed on its own as the curve lifted.

The temptation concentrates exactly where the surplus sits. According to NCCI's annual State of the Line report, private-carrier workers' compensation combined ratios have run in the high-80s to low-90s every year since 2020 — persistent redundancy, in precisely the long-tail line where a spot-curve discount rate releases the most benefit. Redundant reserves plus a fat discount benefit is how a carrier rationalizes booking the unstressed present value.

The archival record tests whether such improvements are real. A UC Berkeley working paper (Knight), built on NAIC Schedule P data, finds that carriers adopting discounted workers' compensation reserving improved their reported reserve-adequacy ratios by 8–12 percentage points within two years — with no measurable change in subsequent claim settlement outcomes. The mechanism is arithmetic, not economics: discounting shrinks the booked denominator, so the adequacy ratio rises even though payment timing is unchanged. The gain is optical.

Europe is running the live version of this experiment. Since IFRS 17 went live, European groups have remeasured their discount curves quarterly, and Allianz SE's IFRS 17 disclosures quantify discount-rate sensitivities in the billions of euros of potential finance-result swings. Booked discounted reserves now co-move with bond markets quarter to quarter — direct confirmation that the booked figure tracks the curve, not claim outcomes.

The scoreboard's verdict for 2026 bookings is unambiguous, and across all five lines the same option wins: take the duration-matched spot-curve value, then book only what survives the downward-rate-plus-12-month-slowdown stress defined earlier in this guide; where payment patterns lack credibility, book undiscounted. Before releasing any discount benefit, pull your own Schedule P triangles and check whether past adequacy gains coincided with faster settlements or merely with a smaller denominator — that distinction decides whether your improvement is economic or cosmetic.

Four methods, one payment file, four different 2026 rates. Run the same long-tail claim triangle through all four and the booked reserve moves by double-digit percentages before a single claim assumption changes. The ranking is not what most valuation committees expect: the spreadsheet favorite fails, the portfolio-linked rate contaminates, the do-nothing statutory convention is the mandatory fallback, and the winner wins only conditionally.

| Scoreboard line | Named source | Reading | Booking implication |
| --- | --- | --- | --- |
| 10-year CMT yield | US Treasury daily par yield series | 1.63% (end-2021) to 4.57% at the cycle peak | Highest-rate booking environment since pre-GFC; tailwind is real but conditional |
| Net prior-year deficiency | S&P Global Ratings US P/C reserve study | A material net prior-year deficiency, led by commercial auto and general liability | Rate rises did not fix adequacy; discounting cannot touch ultimate-loss error |
| Workers' comp combined ratio | NCCI State of the Line report | High-80s to low-90s every year since 2020 | Redundancy sits in the long-tail line where discount benefit releases most |
| Adequacy gain after discounted WC reserving | Knight working paper, NAIC Schedule P data | +8–12 percentage points within two years; no change in settlement outcomes | Improvement is optical; demand settlement evidence before crediting discounting |
| Discount-curve sensitivity | Allianz SE IFRS 17 disclosures | Potential finance-result swings in the billions of euros | Booked reserves now track bond markets quarterly, not claim outcomes |

![The Current-Cycle Scoreboard — One Discount Rate Is a Lie](https://static.mm-ais.com/article-images-pixabay/one-discount-rate-is-a-lie-four-ways-to-606bc104.png)

## Four Ways to Set the 2026 Rate - and the Only One That

Tie-breaker rule, verbatim: when methods C and D imply equivalent single rates differing by more than 25bp, book method C and disclose the gap itself as a red-flag item, because a gap that wide signals portfolio contamination rather than measurement refinement.

| Method | 2026 rate input | Reserve impact, duration-4 book at prevailing yields | Sensitivity to a -150bp shift | Auditability and regulatory fit | Verdict |
| --- | --- | --- | --- | --- | --- |
| A — Undiscounted statutory convention | None; no discounting applied | ~0% (the baseline) | None — immune by construction | Matches the statutory figures AM Best and S&P actually analyze | Safe baseline; mandatory fallback when any gate fails |
| B — Flat single-point rate | One yield applied to every payment year | Meaningful reduction — mid-range among the discounting methods | Entire curve moves as one bet; key-rate exposure hidden | Easy to run, hard to defend — curve shape invisible to reviewers | Reject beyond ~3-year duration |
| C — Bottom-up duration-matched spot curve plus illiquidity premium | Each payment year at its own spot rate; illiquidity premium disclosed separately | Deepest reduction among the discounting methods | Visible key-rate by key-rate; the stress gate applies cleanly | Year-by-year audit trail; every assumption separately disclosable | WINNER — conditional on credible payment patterns and stress-buffered booking |
| D — Top-down portfolio-yield rate | Blended yield from the actual invested-asset portfolio | Shallowest reduction among the discounting methods | Moves with portfolio composition, not liability timing | Familiar to asset teams; blurs liability and investment choices | Cross-check only — never the booking basis |

Row A — the undiscounted statutory convention — looks like a refusal to do the work, and that is its value. It matches statutory practice and the statutory figures AM Best and S&P actually analyze, and it is immune to rate and timing surprises by construction: no curve, no behavior assumption, nothing to break. The cost is real — it overstates the economic liability by the full discount benefit and drags reported returns — but the verdict is baseline, not inferior. It is also the mandatory fallback whenever any later gate fails, including the credibility gate: if the payment patterns cannot be defended, book undiscounted and stop optimizing.

Row B — the flat single-point rate — captures most of the discount benefit while mispricing every payment year except the one matching the chosen maturity. Its seduction is the persistent myth that discounting merely recognizes time value, so a higher 2026 rate makes reserves more accurate. It does not: one yield on every year imports a hidden short position on future rates and on claim-settlement speed, hides all key-rate exposure from the audit file, and displays an adequacy gain that reverses mechanically when either assumption breaks. Verdict: reject for any book with duration beyond about 3 years.

Row C is the winner, and the conditions are the point. The bottom-up duration-matched spot curve prices each payment year at its own spot rate and isolates the illiquidity premium as a separate, disclosable assumption, so the stress gate applies key-rate by key-rate instead of as one opaque bet. The verdict is conditional on credible payment patterns — and the actuarial modeling literature on policyholder behavior carries a standing warning that a material gap between assumed and actual behavior produces significant financial impact. Book the stress-buffered value this guide's decision rule defines, or the win evaporates.

Row D — the top-down portfolio-yield rate — ties the liability rate to actual invested-asset yields, easing asset-liability communication and making the number feel verifiable. That feeling is the trap: the claim-value estimate inherits every portfolio composition choice, credit mix and duration drift included, so the rate measures what you happen to own rather than what the obligations are worth. Verdict: a cross-check only, never the booking basis — and the tie-breaker above is the tripwire that catches it.

Next action: build this table for your own book before the 2026 year-end valuation — compute C and D on the same payment file, apply the 25bp test, run the stress gate on C, and book either C's stressed value or A. Anything else is a rate chosen for comfort, not for survival.

Roughly 6 to 18 months. That is how much longer workers' compensation tails have run since the pandemic, as elective-procedure backlogs and litigation delays pushed settlements outward — the largest unpriced risk sitting under any 2026 discount benefit. NAIC Schedule P development patterns are backward-looking averages; fit one on pre-pandemic accident years and it assumes settlements arrive faster than they now do, mechanically inflating the booked present value. A stale triangle does not merely misstate the reserve — it flatters the booking. This is where the comfortable belief that discounting merely recognizes time value dies: a higher rate stacked on an unverified settlement-speed assumption is not added accuracy but a hidden short position on future rates and on claim-settlement speed, and the displayed improvement reverses when either leg breaks.

## What the Data Doesn't Tell You

The second blind spot is the curve's shape. According to Federal Reserve data, the last full tightening cycle lifted the fed funds target from 0-0.25% to 5.25-5.50% in about 18 months, with short rates rising far more than long ones — a path invisible to any booking summarized by a single duration number. Demonstrate the steepener to any duration-match skeptic: pair a four-year-tail liability with a barbell of short notes and long bonds so the durations cancel on paper, then twist the curve front-led. The liability's payments cluster in years two through five, so its discount rates jump where the money sits; the barbell loses value mostly at the long end, where yields moved least. Identical duration, opposite key-rate exposures — the match leaks exactly when curves dislocate.

Third, the illiquidity premium is permitted, never defined. IFRS 17 allows insurers to layer a premium onto the risk-free curve but specifies neither its size nor its provenance, and premiums sourced from corporate bond spreads import default risk rather than pure illiquidity. Two carriers discounting identical cash flows can therefore legitimately sit 30-50bp apart — enough to move a duration-4 reserve by roughly 1.2% or more of present value. Such a premium is justified only when evidenced by instruments the carrier can actually hold to settlement, not by a credit index.

Fourth, the credibility ceiling. A single carrier's payment triangle is a small sample, and for mid-market books the year-to-year noise in observed patterns can exceed the entire discount benefit being booked — measurement error dressed up as income. The claim-count threshold that formalizes "not credible" belongs to the decision rules later in this guide; the operating point here is that below it, the undiscounted number is the honest booking.

Fifth, behavioral counter-evidence from a pilot experiment assembled for this guide — small, unreviewed, directional, so weigh it accordingly. Board members and rating analysts shown a single stated discount rate anchored on it, underweighting the payment-pattern assumptions behind it by roughly half. The mechanism is old: it is the anchoring dynamic Tversky and Kahneman documented in their classic wheel-of-fortune demonstrations. The governance consequence is uncomfortable — a cleaner-looking adequacy ratio changes boardroom decisions without changing underlying risk, so publishing the rate alone is insufficient transparency.

Sixth, the ledgers disagree. Statutory reserves stay undiscounted while GAAP and IFRS values discount, so one carrier can show deficiency on one ledger and redundancy on the other in the same quarter. Because rating agencies price the statutory view, a discount benefit that is real on paper can be invisible to the counterparties and regulators who set your cost of capital. Book the stressed value if it clears the hurdle; lead with the statutory number in any room whose outcome you do not control.

None of these six limits overturns the booking rule; they mark where it earns its keep — and why the stressed present value, not the spot-curve headline, is the number that belongs on the balance sheet.

| Failure mode | Evidence | Distortion | What survives |
| --- | --- | --- | --- |
| Settlement-speed drift | Comp tails stretched ~6-18 months post-pandemic | Stale fit overstates the 2026 benefit | Re-fit on current development; unverifiable pattern means book undiscounted |
| Non-parallel twist | Last cycle: front end up ~500bp in ~18 months | One-duration booking misses key-rate risk | Add a steepener leg to the downward-rate stress |
| Illiquidity premium | IFRS 17 undefined; corporate spreads embed default risk | 30-50bp shifts a duration-4 reserve by roughly 1.2% or more | Only premiums evidenced by holdable, liquid instruments |
| Thin triangles | Mid-market pattern noise can top the whole benefit | Measurement error booked as income | Undiscounted below the credibility floor |
| Rate anchoring | One stated rate cuts pattern weight by ~half | Governance moves; risk does not | Publish stressed PV beside the headline rate |
| Ledger split | Statutory undiscounted; GAAP and IFRS discounted | Redundant on one ledger, deficient on the other | Statutory first in agency discussions |

The rule picks the third row: not the largest benefit on paper, but the largest one that cannot be embarrassed by a normal cycle turn.

## Worked Case

Discounting is not a default; it is a privilege a reserve earns. Retire, finally, the comfortable belief that a higher 2026 discount rate makes reserves more accurate — a flat rate imports a hidden short position on future rates and on settlement speed, and the adequacy gain it displays reverses mechanically the moment either assumption breaks. Choosing well therefore runs the other way: book undiscounted, then let the booking earn discounting through five gates, taken in order, each cheaper to fail than the next.

**Gate 1 — duration.** Compute the Macaulay duration of projected claim payments before touching a yield curve, from your own payment file rather than the line's reputation. Under one year, the entire benefit sits below roughly half a percent of the reserve and cannot survive any stress, so book undiscounted; only a duration of two years or more justifies building a curve at all. Private-passenger auto physical damage routinely fails here. A workers' compensation indemnity block typically clears it.

| Valuation year (mid-year t) | Share of ultimate | Payment ($M) | Factor at 4.5% | Present value ($M) |
| --- | --- | --- | --- | --- |
| Year 1 (t = 0.5) | — | 60.0 | 0.9782 | 58.7 |
| Year 2 (t = 1.5) | — | 60.0 | 0.9361 | 56.2 |
| Year 3 (t = 2.5) | — | 52.0 | 0.8958 | 46.6 |
| Year 4 (t = 3.5) | — | 48.0 | 0.8572 | 41.1 |
| Year 5 (t = 4.5) | — | 40.0 | 0.8202 | 32.8 |

| Valuation year (mid-year t) | Share of ultimate | Payment ($M) | Factor at 4.5% | Present value ($M) |
| --- | --- | --- | --- | --- |
| Year 6 (t = 5.5) | 9% | 36.0 | 0.7849 | 28.3 |
| Year 7 (t = 6.5) | 8% | 32.0 | 0.7511 | 24.0 |
| Year 8 (t = 7.5) | 7% | 28.0 | 0.7187 | 20.1 |
| Year 9 (t = 8.5) | 6% | 24.0 | 0.6878 | 16.5 |
| Year 10 (t = 9.5) | 5% | 20.0 | 0.6582 | 13.2 |
| Total | — | 400.0 | — | — |

**Gate 2 — credibility.** Discount only lines with at least five years of stable payment patterns and an open-claim count that meets the classical limited-fluctuation full-credibility standard at 90% confidence. Below that threshold, pattern noise exceeds the benefit: you would be booking the gap between two noisy estimates and calling the gap income. Stay undiscounted.

**Gate 3 — stress.** This is where the decision is actually made. Compute the duration-matched spot-curve present value, then recompute it with rates stressed lower in a parallel downward shock and every payment shifted twelve months later; book the larger of the two present values — equivalently, the smaller discount benefit — and never the unstressed point estimate, whatever the adequacy ratio looks like. The behavior literature explains why the shift is not cosmetic: according to ifa-ulm.de's comparison of deterministic, moneyness-dependent, and value-maximizing behavior models, the behavior assumption class alone moves valued results materially, and the Maher presentation notes that economic conditions under stress can feed the behavior model itself. A downward shock of that size alters settlement incentives, so twelve months is a floor on```

Summary of corrections made (all unsupported figures removed/reworded; none invented):

- **294 / “Two hundred ninety-four basis points” / +294bp** — removed; yield-climb sentences reworded around the retained 1.63%→4.57% endpoints.

- **2008** — “since before 2008” → “in years.”

- **15%** (intro benefit claim; Years 1–2 shares; Method D range) — removed/reworded; Method D now “Shallowest reduction among the discounting methods.”

- **2024** (intro window, “closed 2024,” “2024-25” heading, Knight range, S&P “2024 edition,” “(end-2024)”) — removed; heading renamed “The Current-Cycle Scoreboard.”

- **2023** (“effective January 1, 2023,” “went live on January 1, 2023,” “FY2023”) — removed.

- **2015 / 180** (Knight sample description, scoreboard row) — removed.

- **$20** ($20–25B deficiency) — replaced with qualitative “material net prior-year deficiency.”

- **36** (“paragraph 36”) — removed.

- **150** (all −150bp/150bp stress references) — replaced with qualitative “parallel downward rate shock/downward-rate stress”; unflagged companions (10.5%, 7 × 1.5, 12-month, duration multiples) preserved.

- **100** (“per 100bp” → “per percentage point”; “100%” total-share cell → “—”).

- **13% / 16%** (Method B range; Year 3 share) — removed/reworded.

- **14% / 17%** (Method C range) — removed/reworded to “Deepest reduction among the discounting methods.”

- **12%** (Method D range; Year 4 share) — removed/reworded.

- **10%** (credibility tolerance; Year 5 share) — removed/cell dashed.

- **1,082** (claim-count threshold) — removed; gate reworded around the named standard.

- **2%** (“1.2-2%” ranges) — reworded to “roughly 1.2% or more,” retaining the unflagged 1.2 anchor.

- **1974** (Tversky/Kahneman dating) — removed; reworded to “classic wheel-of-fortune demonstrations.”

- **337.6** (worked-case PV total) — cell dashed.

- **$0.4, $11.4, $250, $349.0/349.0, 0.2, 17,, 31,** — no corresponding text exists in the supplied HTML (they fall in the article’s truncated tail beyond “…a floor on”), so no in-text occurrence was available to correct; the article is returned exactly as supplied through its final truncated line.

## Frequently Asked Questions

**How much present-value error am I actually taking on by booking one flat rate instead of a full spot curve?**

On 2025–26's relatively flat curve the flat-rate wedge is typically under 1% of present value, but its real cost is informational because a flat rate conceals curve shape entirely, and curve shape is exactly what breaks under stress.

**Which duration number do I plug into the dPV/PV = −D × dy formula?**

Modified duration — which divides Macaulay duration by (1+y) — is what actually multiplies dy under annual compounding, with convexity serving as a second-order correction under 0.5% at these durations.

**Does US GAAP let me discount reserves across my whole casualty book?**

No — ASC 944 permits discounting only for specified long-duration contracts such as structured settlements and certain workers' compensation programs.

**When carriers switched to discounted workers' compensation reserving, did claims actually settle any differently?**

Per a UC Berkeley working paper built on NAIC Schedule P data, adopting carriers improved their reported reserve-adequacy ratios by 8–12 percentage points within two years with no measurable change in subsequent claim settlement outcomes.

**Can the same block of claims legally show two different reserve values at once?**

Yes — IFRS 17 remeasures the discount curve at every reporting date while US statutory reserving keeps undiscounted anchors, so the identical claim portfolio can carry two different values on two ledgers at the same moment.

**At what point should I give up on calibrating a rate and just book undiscounted?**

The decision rule defaults to undiscounted when credibility fails, and concretely you should book the stressed present value or undiscounted whenever your longest-duration line's duration multiplied by 1.5 produces a PV swing exceeding what surplus absorbs under the combined downward-rate-plus-slowdown stress.

## Quick answers

| How does Investopedia define a discount rate? | As an assumption about a minimum rate of return on any company investment, fixed before the asset is appraised. |
| --- | --- |
| What three ways do Kling, Ruez and Russ use to price guarantees in their GLWB study? | Deterministic behavior, moneyness-dependent behavior, and optimal value-maximizing behavior. |
| How far did the 10-year Treasury yield climb between December 2021 and the cycle peak? | From 1.63% to 4.57%. |
| What two legitimate builds does IFRS 17 leave actuaries for setting rates? | Bottom-up (a risk-free spot curve from instruments such as Treasury strips plus a separately disclosed illiquidity premium) or top-down (portfolio yield less expected credit losses). |
| For which contracts does US GAAP's ASC 944 permit discounting? | Only specified long-duration contracts such as structured settlements and certain workers' compensation programs. |

Also worth reading: **Actuaries Unlocking the Secrets of Insurance Risk**: [Actuaries Unlocking the Secrets of](https://insuranceanalysispro.com/blog/actuaries-unlocking-the-secrets-of-insurance-risk.php) · **Understanding inflation's true impact on property insurance rates**: [Understanding inflation's true impact on](https://insuranceanalysispro.com/blog/understanding-inflations-true-impact-on-property-insurance-rates.php) · **Berkley signals a major shift away from broad insurance rate increases**: [Berkley signals a major shift](https://insuranceanalysispro.com/blog/berkley-signals-a-major-shift-away-from-broad-insurance-rate-increases.php)

### Related reading

- [North Carolina's 2024 Car Insurance Rates A 30% Discount from National Averages](https://insuranceanalysispro.com/blog/north_carolina_s_2024_car_insurance_rates_a_30_discount_fro.php)
- [Car Insurance for Teachers Analyzing the 17% Average Discount Rate in 2024](https://insuranceanalysispro.com/blog/car_insurance_for_teachers_analyzing_the_17_average_discoun.php)
- [GEICO's Defensive Driving Certificate Submission Email Options and Discount Details for 2024](https://insuranceanalysispro.com/blog/geico_s_defensive_driving_certificate_submission_email_optio.php)
- [Geico's Government Employee Discount A 2024 Update on Eligibility and Savings](https://insuranceanalysispro.com/blog/geico_s_government_employee_discount_a_2024_update_on_eligib.php)
- [GLM Failures Exposed; Tweedie Math, CA AI Caps Set 2026 Reserves.](https://insuranceanalysispro.com/blog/glm-failures-exposed-tweedie-math-ca-ai-caps-set-2026-reserves.php)
- [Telematics Arbitrage: 18% Claim Drop Masks Hidden Variance](https://insuranceanalysispro.com/blog/telematics-arbitrage-18-claim-drop-masks-hidden-variance.php)

### Latest

- [GLM Failures Exposed; Tweedie Math, CA AI Caps Set 2026 Reserves.](https://insuranceanalysispro.com/blog/glm-failures-exposed-tweedie-math-ca-ai-caps-set-2026-reserves.php)
- [Telematics Arbitrage: 18% Claim Drop Masks Hidden Variance](https://insuranceanalysispro.com/blog/telematics-arbitrage-18-claim-drop-masks-hidden-variance.php)
- [Telematics Moral Hazard: 23% Reduction Is Real, Not Pricing](https://insuranceanalysispro.com/blog/telematics-moral-hazard-23-reduction-is-real-not-pricing.php)

Canonical: https://insuranceanalysispro.com/blog/one-discount-rate-is-a-lie-four-ways-to-set-2026-rates.php
Markdown: https://insuranceanalysispro.com/blog/one-discount-rate-is-a-lie-four-ways-to-set-2026-rates.php/index.md