What Discount Rate Belongs on Social Security?

Every argument about when to claim Social Security is an argument about one number that the argument rarely states. Claiming at 62 gives you 70% of your full benefit for more years. Waiting until 70 gives you 124% for fewer. Comparing those two streams means discounting future dollars back to today, and the rate you pick decides the answer before any other fact about your life gets a vote.

The rate is the decision, and it needs a date attached, because it moves.

What the statute gives you

The trade-off is fixed in law, not set by a market. For anyone born in 1960 or later, full retirement age is 67. Claiming early costs 5/9 of 1% a month for the first 36 months and 5/12 of 1% a month beyond that, so claiming at 62 means a 30% reduction. Delaying past full retirement age earns 2/3 of 1% a month, or 8% a year, until it stops at 70.¹ That leaves three reference points: 70% of the full benefit at 62, 100% at 67, and 124% at 70.

One caveat: the 5/9 of 1% formula is the retired-worker rate. Spousal benefits are reduced by 25/36 of 1% a month for the first 36 months and 5/12 of 1% beyond, so a spouse claiming at 62 against a full retirement age of 67 receives 65% of the spousal amount, and spousal benefits earn no delayed retirement credit at all.¹

Why the rate decides it

Social Security pays an amount set by statute, adjusted for inflation every year, for as long as you live. A Congressional Research Service report on the claiming adjustment describes how researchers treat a cash flow with those properties, hedge included: “There is no clear conclusion on what interest rates should be used when discounting future Social Security benefits. Many researchers agree that the government-guaranteed, tax preferred, inflation-indexed nature of the Social Security benefit warrants a low discount rate.”²

An article published in SSA’s own Social Security Bulletin states the mechanism and the consequence. On the mechanism: the studies it reviews share the view that “the single discount rate should be related to the yield on long-term inflation-indexed government bonds because Social Security benefits are also inflation-adjusted financial obligations of the U.S. government and should thus be valued similarly.” On the consequence: “Viewed prospectively, the optimal claiming age is older at lower discount rates and younger at higher rates.”³

The numbers, with real mortality

Here is the present value at age 62 of each claiming strategy, per dollar of full retirement age benefit, with every future year weighted by the probability of being alive to collect it. Mortality comes from SSA’s 2023 period life table as used in the 2026 Trustees Report, so no fixed death age is assumed.

Male mortality, scheduled benefits, present value at 62 per $1 of annual full retirement age benefit. Our calculation; the script and life table are published at github.com/engineerinvestor/summitward-research.

Women live longer, so the same table moves. At 2% real, the ranking flips back to delaying:

Two crossovers come out of those tables, and they are different numbers. The highest real rate at which waiting to 70 still beats claiming at 62 is 2.3% for a man and 3.4% for a woman on this model. The highest rate at which 70 also beats 67, so that 70 has the largest present value of the three, is 1.4% and 2.7%. Between the two, 67 is the best age; the male table shows it at 2%, where 67 leads both neighbours. We quote one decimal because switching the model from annual to monthly payments moves each figure by about 0.15 points, and a period life table describes a population rather than any one person.

Shoven and Slavov, using cohort mortality, find that delay is actuarially advantageous for a large subset of people “particularly for real interest rates of 3.5 percent or below,” and that “in a two-earner couple, the gains from deferring the primary earner’s benefit are greater than the gains from deferring the secondary earner’s benefit.”⁴ Their threshold sits above ours because a cohort table expects a 62-year-old to live longer than the 2023 period table does, and the readers of a personal finance site, who are on average higher earners, live longer still. Every one of those adjustments raises the crossover and strengthens the case for waiting.

Move the rate yourself

The calculator starts at 6% real. Drag it down toward what an inflation-protected Treasury yields today and watch the ranking change.

What today’s real yields say

The instrument whose risk most resembles a Social Security benefit is a long inflation-protected Treasury, and a 62-year-old’s benefit stream has a duration closer to a 20-year bond than a 10-year one. On September 10, 2026 the 20-year TIPS yield was 2.87% real and the 30-year 3.05%, with the 10-year at 2.55%, per our markets page. Mike Piper’s Open Social Security calculator, the standard free tool for this decision, uses the 20-year TIPS yield as its default discount rate for the same reason.⁵

At 3% real, which is where the long end of the curve sat when this was written, our model puts a man’s present values at 10.51, 10.43 and 10.10 for 62, 67 and 70, so claiming at 62 is highest by about 4%. For a woman the same row reads 11.58, 11.90 and 11.82, so 67 is highest by about 3%. Those are small margins on a model with known biases toward early claiming, and they are outputs of one population table applied to one date’s yields. A year earlier the 20-year yield was 2.22%, and the male ranking would have read 67 first. The direction of the conclusion is stable: at any long TIPS yield of the last two years, the three ages are within a few percent of each other and the choice belongs to the factors below.

What a trust-fund shortfall does to the ranking

The 2026 Trustees Report projects that the retirement trust fund is depleted in the fourth quarter of 2032, after which incoming taxes cover 78% of scheduled benefits, and that the combined retirement and disability funds last until 2034, with 83% payable.⁶ Congress has closed every previous shortfall before it arrived, most recently in 1983, and the benefit is a statutory entitlement rather than a contract, so the honest label for the cash flow is “inflation-indexed and set by law” rather than “guaranteed.”

A cut that arrives on a date, rather than at claiming, does not treat the three ages alike. Someone who is 62 in 2026 and claims now collects about seven years of unreduced benefit before 2032; someone who waits to 70 starts collecting after the cut and never receives an unreduced dollar. Rerunning the male table with 78% of the scheduled benefit paid from age 69 on:

Male mortality, 78% of the scheduled benefit paid from age 69, present value at 62 per $1 of annual full retirement age benefit. The 70-over-62 crossover falls to 0.5% for a man and 1.7% for a woman. Same script, with the haircut option.

That is the strongest arithmetic argument for claiming early, and it is the one the April 2026 article did not make. It also depends on Congress doing nothing for six more years, which has no precedent. A reader who expects the shortfall to be closed the way the 1983 one was should read the scheduled-benefit tables; a reader who expects the cut should read this one; most people belong somewhere between.

How this goes wrong in public

A well-read example arrived in April 2026, when Mr. Money Mustache published an analysis of Social Security for early retirees. The observation behind it is good: people who retire at 40 systematically undercount a benefit that starts arriving at 62, and it does have value today. The problem is the price he puts on it.

The rate appears once, in an assumption list, and is never revisited or tested: “A compounding rate for investments of about 6% after inflation.” From there the conclusion is quick. “By the time your more patient friends started drawing $2796 payments five years later, you’d already have over $137,000. It’s such a big lead that the 67-year-old will never catch up.”⁷ At 6% real, that is exactly what the arithmetic says. The tables above show why, and they show that the size of the lead belongs to the assumption rather than to Social Security. At the 2.9% a 20-year TIPS paid when this was written, the model still leans toward 62 for a man, by about 4% rather than by a margin nobody could catch up to, and it leans toward 67 for a woman.

The article also fixes the date of death, at 92 for its couples and 90 for its single example, instead of weighting each year by the chance of reaching it. That deletes the right tail of the longevity distribution, which is where the whole value of waiting lives: the person who waited holds a smaller account balance and a larger inflation-indexed annuity that pays for life, and “never catch up” compares only the balances.

A smaller inconsistency runs the other way. The article prices the benefit as the lump sum that would replicate it over a fixed term: “if they invested $834,000 into a fund that paid out 6% per year for 30 years before running dry, they’d get that exact same $5000 monthly income.” For its single example it then adds the equivalent figure to the portfolio and applies a perpetual rule: “If you take this total amount ($460,000) and apply my other shockingly simple math number, the 4% rule, you end up with $18,400.”⁷ A present value that amortises its principal over 30 years is being spent under a rule that assumes the principal survives. The two conventions do not fit together, and in this case the mismatch is conservative: 4% of a $210,000 present value is $8,400 a year, against a benefit stream of about $23,600. It understates his example’s Social Security, so it does not help his conclusion.

Two things the article leaves out are material to its own recommendation: spousal and survivor benefits, and the taxation of benefits. The retirement earnings test is not among them. The article’s examples are retirees, and when a reader asked about a partner still in a well-paid job, he answered in the comments that “there is no real benefit to drawing the social security early.”⁷

The discount-rate objection was also made to him directly, on the day of publication. A reader, Jesse Cramer, commented: “Please consider using a discount rate that meets the criteria of the ‘next best option of similar risk.’ That is standard practice for discount rates. In this case, TIPS would be a natural choice. Their inflation-adjusted discount rate would be ~2%. ... Using the proper 2% instead of 6% would lead us to the opposite conclusion as this article.” He replied the next morning: “It’s a fair point but I happen to disagree. Index funds are more volatile and not guaranteed, but not actually more ‘risky’ in terms of the return you should forecast over a multi decade period.”⁷

That reply is the disagreement in one sentence, and it conflates two different quantities. The expected return on stocks over thirty years is indeed around 6% real. The rate at which a riskless, inflation-indexed stream should be valued is the rate a riskless, inflation-indexed instrument pays, because the extra 3 to 4 points on equities are compensation for the chance that the thirty years go badly, and discounting a stream that carries none of that chance at a rate that is mostly payment for it throws the payment away. By the same reasoning every pension, annuity and TIPS ladder would be worth less than it costs. Where he is right is the direction: at today’s long TIPS yields, which are well above the 2% in that comment, claiming at 62 is close to or slightly ahead of waiting for a man on our model, and the mistake is the size of the claimed lead rather than its sign. The broader audit of that post is in Fact-Checking Mr. Money Mustache.

When a higher personal rate is defensible

The case above is about computing the expected present value of a benefit stream. It is not a rule that every person must apply a Treasury yield to their own decision, and the SSA-published article is explicit about this. Alleva notes that while some experts hold that the only appropriate rate is the long-term government bond yield, “a more appropriate rate choice would reflect the particular needs of a given individual.”³ Someone servicing expensive debt, or facing a real liquidity constraint at 62, is rationally impatient, and impatience is a high discount rate. So is someone who would have to sell stocks in a drawdown to bridge the years from 62 to 70: the money that funds a delay comes out of the portfolio during the years in which sequence risk is highest, and a present-value table does not price that.

There is also a fair reading in which the whole exercise matters less than it appears. Alicia Munnell’s work at Boston College concludes that because the delayed retirement credit has grown, “lifetime benefits are roughly equal for any claiming age between 62 and 70, and the highest monthly benefits are available at 70.”⁸ Our tables agree at realistic rates: at 2% real the three present values sit within 3% of each other for a man and 8% for a woman. The case for delay at that point rests on insurance rather than arbitrage. You are buying a larger inflation-protected income for the years in which running out of money would hurt most. Munnell also names the ways the benefit is less than fully indexed in practice, which the tables ignore: Medicare premiums are deducted from it and have grown faster than the cost-of-living adjustment, the adjustment itself follows CPI-W, and the income thresholds above which benefits become taxable have never been indexed, so a rising share of the benefit is taxed over time. The standard treatment of the claiming decision as a present-value problem, with these factors folded in, is Meyer and Reichenstein’s.⁹

What we recommend

Price the benefit at a rate that matches its risk, which means the 20-year TIPS yield on the day you look, rather than an equity return. Do that first, before any argument about investing the early payments, because the discount rate is the assumption that decides the outcome, and write the date next to it, because at 2% the model says wait and at 3% it says do not.

Then treat the result as close and let the non-arithmetic factors break the tie. Delay looks strongest for the higher earner in a couple, because that benefit becomes the survivor benefit, and for anyone whose family history or health suggests the right tail. It looks weakest for someone in poor health, someone with no other assets to bridge the gap, someone still working before full retirement age, and anyone who expects the 2032 shortfall to arrive unrepaired. If you want the break-even framing and the taxation of benefits in full, that is in When to Claim Social Security.

Frequently asked questions

What discount rate should I use for Social Security?

For computing the present value of the benefit itself, the 20-year TIPS yield, because that is the instrument whose risk and duration most resemble the benefit; Open Social Security uses the same default. On September 10, 2026 it was 2.87% real, up from about 2.2% a year earlier. A higher personal rate can be defensible if you are genuinely liquidity-constrained, but it is a statement about your circumstances rather than about the asset.

Does claiming at 62 and investing the money beat waiting?

On our model, claiming at 62 has a higher present value than waiting to 70 at any real rate above about 2.3% for a man and 3.4% for a woman, which includes the long TIPS yields of late 2026, and the margins are a few percent either way. The argument for claiming early and investing usually assumes the early payments earn an equity return with certainty, which substitutes a risky return for a riskless one without pricing the difference; the version that survives is the one that discounts at the TIPS yield and accepts a small edge rather than a landslide.

Is Social Security a bond or an annuity?

It behaves like an inflation-indexed life annuity paid by the federal government under a statute Congress can change, which is why a real government bond yield is the natural starting point for valuing it and why a haircut for the projected 2032 shortfall is a reasonable second step. How that should change the rest of your portfolio is a separate question, covered in Should Retirees With Social Security or a Pension Hold More Stocks?

Do these tables apply to spousal benefits?

No. They model retired-worker benefits. Spousal benefits use a different reduction factor, 25/36 of 1% a month for the first 36 months, and earn no delayed retirement credit, so waiting past full retirement age adds nothing to a spousal benefit.

What about future benefit cuts?

Modelled above. The 2026 Trustees Report projects 78% of scheduled retirement benefits payable from late 2032 if Congress does nothing. A cut that arrives on a date favours the early claimer, who collects several unreduced years first, over the person who waits and never collects one. On our model that alone pulls the rate below which waiting to 70 beats 62 from 2.3% to 0.5% for a man. Congress has closed every previous shortfall before it arrived, so treat the cut as a real possibility rather than a certainty, and as an argument for earlier claiming rather than a reason to ignore the question.

Key takeaways

• The discount rate decides the claiming answer. Pick it before you argue about anything else, and date it, because between 2% and 3% real the ranking of the three ages reverses.
• On our model of SSA’s factors and 2023 life table, waiting to 70 beats 62 below about 2.3% real for a man and 3.4% for a woman. 70 has the highest present value of the three ages only below about 1.4% and 2.7%; between those rates, 67 wins.
• Match the rate to the risk of the cash flow. Social Security is an inflation-indexed life annuity set by statute, so a long real Treasury yield is the natural comparison. On September 10, 2026 the 20-year TIPS yield was 2.87%, which favours 62 for a man and 67 for a woman on this model, by a few percent.
• A dated benefit cut favours early claiming. With 78% of scheduled benefits paid from late 2032, as the Trustees project if Congress does nothing, the rate below which waiting to 70 beats 62 falls to 0.5% for a man.
• At realistic rates the three options are close. Munnell finds lifetime benefits roughly equal across 62 to 70, which moves the decision onto survivor benefits, longevity, health, employment and the sequence risk of bridging the gap.
• The statutory factors are not symmetric across benefit types. Spousal benefits use a different reduction and earn no delayed credit.

Related guides

• When to Claim Social Security the break-even math, three case studies, and why up to 85% of benefits can be taxable.
• Should Retirees With Social Security or a Pension Hold More Stocks? what a guaranteed income stream does to the rest of the portfolio.
• Fact-Checking Mr. Money Mustache the wider audit that this analysis came out of.
• How Much of Your Retirement Should Be a Guaranteed Income Floor? where a delayed benefit fits in a floor-and-upside plan.

Author disclosure

Summitward sells a personal finance product and publishes a retirement planner, so we have an interest in readers believing this decision is worth modelling carefully. We have tried to state the case against our own conclusion as clearly as the case for it, including the finding that at realistic rates the three claiming ages are nearly tied. Nothing here is personalized financial advice.

Sources

1. 20 C.F.R. §§ 404.313 (delayed retirement credit, 2/3 of 1% per month for those born after January 1, 1943), 404.409 (full retirement age table; 67 for those born January 2, 1960 and later) and 404.410 (reduction of 5/9 of 1% for the first 36 months and 5/12 of 1% beyond, with 25/36 of 1% for spouses). ecfr.gov
2. Congressional Research Service, Report R47151, Social Security: Adjustment Factors for Early or Delayed Benefit Claiming. everycrsreport.com
3. Alleva, B.J. (2016). “Discount Rate Specification and the Social Security Claiming Decision.” Social Security Bulletin 76(2). Office of Retirement Policy, Social Security Administration. Alleva argues against treating a bond yield as the only admissible rate, while confirming the direction of the dependence. ssa.gov
4. Shoven, J.B. and Slavov, S.N. (2012). “The Decision to Delay Social Security Benefits: Theory and Evidence.” NBER Working Paper 17866. Published with its companion as “Does it pay to delay social security?” Journal of Pension Economics & Finance 13(2), 2014, 121–144. nber.org
5. Piper, M., Open Social Security. The tool’s help text states that its default discount rate is the yield on 20-year TIPS, fetched from the Treasury daily. Real yields quoted in this guide are from Treasury’s daily par real yield curve as carried on our markets page, observation dated September 10, 2026. opensocialsecurity.com
6. Social Security Board of Trustees, press release, June 9, 2026: OASI reserves depleted in the fourth quarter of 2032 with 78 percent of benefits payable; combined OASDI reserves depleted in 2034 with 83 percent payable. ssa.gov
7. Mr. Money Mustache, “The Shockingly Simple Math Behind Social Security,” April 16, 2026. Read September 14, 2026. The discount-rate exchange is in the comments: Jesse Cramer, April 16, 2026, 8:17 pm, and the author’s reply, April 17, 2026, 8:42 am. The earnings remark is the author’s reply to a reader, April 16, 2026, 3:53 pm. mrmoneymustache.com
8. Munnell, A.H. (2013). “Social Security’s Real Retirement Age Is 70.” Center for Retirement Research at Boston College, Issue in Brief 13-15. crr.bc.edu
9. Meyer, W. and Reichenstein, W. (2010). “Social Security: When to Start Benefits and How to Minimize Longevity Risk.” Journal of Financial Planning 23(3), 49–59. The present-value framework for singles and couples, including survivor benefits, on which most later claiming-strategy work builds.
10. Social Security Administration, Office of the Chief Actuary, period life table for 2023 as used in the 2026 Trustees Report. ssa.gov

Editor’s note

Educational content, not financial advice. The present values here are outputs of our own model, published with its life table at github.com/engineerinvestor/summitward-research, applied to the statutory benefit factors and SSA’s 2023 period life table. They assume benefits are fully inflation-indexed and paid once a year in advance; paying monthly lowers each crossover by about 0.15 points. They exclude taxation of benefits, spousal and survivor benefits, and the retirement earnings test, and they model a benefit cut only in the one scenario shown. A period life table describes mortality in one year rather than the cohort experience of someone alive today, and it describes the whole population rather than the higher-earning, longer-lived readers of a site like this one; both understate how long a 62-year-old reader will live and therefore understate the case for delay. Real yields are as of September 10, 2026 and will have moved. Quotations from mrmoneymustache.com were read on September 14, 2026.