Work One More Year or Save More: Which Moves Retirement Success More? (2026)

Work one more year or save more: which moves retirement success more?

Key numbers

work longer save more spend less — persona A: $800,000, $60,000/yr; persona B: $600,000, $67,000/yr.

10,000 quasi-Monte Carlo paths per arm, identical market draws for every arm of a persona; JPMorgan LTCMA 2026 capital-market assumptions (as of 2025-11-01); all dollars in today’s dollars; pre-tax. “Success” means the portfolio never reached $0 before 95. “Per unit” is the success gain divided by years worked, $10,000-a-year saving steps or $5,000-a-year spending steps. Every arm adds money or removes withdrawals on the same draws, so no arm ends below its baseline on any path; “arm ends higher” falls short of 100% only where both arms run out.

Run persona A in the free planner →

Opens the Monte Carlo planner prefilled with this study’s baseline — age 60, retire at 65, plan to 95, $800,000, $30,000 a year saved, $60,000 a year spent, $30,000 Social Security from 67, 45/15/40 US equity / international / bonds. Change the retirement age, the contribution or the spending to see each lever on your own numbers.

What is being compared

The baseline is a 60-year-old who keeps saving $30,000 a year until 65, then draws a constant-real $60,000 a year (persona A) or $67,000 (persona B) from the portfolio, with $30,000 a year of Social Security arriving at 67 and reducing the draw from then on, to 95. The portfolio is 45/15/40 US equity / international / bonds, rebalanced once a year. Six arms change exactly one thing each:

• Work longer. Retire at 66 or at 67: contributions continue at $30,000 a year for one or two more years, withdrawals start one or two years later, spending is unchanged, and Social Security still starts at 67.
• Save more. Retire at 65 but contribute $40,000 or $50,000 a year for the five working years — $10,000 or $20,000 a year more, $50,000 or $100,000 in total before any return.
• Spend less. Retire at 65 and spend $5,000 or $10,000 a year less, in today’s dollars, in every year of retirement.

Every arm of a persona runs on the same 10,000 market paths, so every statistic can be taken path by path. The headline for each arm is the change in the success rate and, to make the three levers comparable, that change per unit: per year worked, per $10,000 a year saved, per $5,000 a year cut. Beside it is the paired difference in wealth at 95 — the arm’s balance minus the baseline’s on the same path — summarised by its median and its 10th–90th percentile band. Because each arm only adds money or removes withdrawals, it ends at or above the baseline on every path; the share of paths on which it ends strictly higher is therefore one minus the share on which both arms run out, and the informative columns are the size of the gap, not its sign.

Results: persona A, $800,000 spending $60,000 a year (comfortable plan)

Baseline success 97.3%. Each row is one arm against the same baseline. Dollar columns are wealth at 95 in today’s dollars.

“Per unit” = Δ success ÷ units (1 or 2 years worked; 1 or 2 steps of $10,000 a year saved; 1 or 2 steps of $5,000 a year cut). “Paired median Δ” is the median over 10,000 paths of (arm wealth − baseline wealth) at 95; the band is the 10th and 90th percentile of that per-path difference. The last column counts paths on which the arm reached 95 with money left and the baseline did not; the reverse never happens.

Results: persona B, $600,000 spending $67,000 a year (marginal plan)

Baseline success 70.9%: the plan is close to the line, and the same six moves now rescue many more paths.

The gap over time

The engine reports each path’s wealth year by year, so the paired difference can be followed from the retirement date onwards. The tables show the median per-path gap (arm minus baseline, today’s dollars) at six ages and the share of paths on which the arm is ahead at that age. The work arms open their gap in the first year of delay; the saving arms have theirs in place by 65 and then let it compound; the spending cuts start from nothing and widen every year.

Persona A

Persona B

Median of the per-path difference at each age, with the share of paths on which the arm is ahead. Ages are year-end balances; age 66 is the end of the first year after the baseline retires.

Why a year of work moves success more than the same dollars saved

A year of delay does three things at once, and only one of them is what “saving more” does. First, it adds a year of contributions: $30,000 more goes in. Second, it removes a year of withdrawals: the $60,000 that would have come out at 65 — the full amount, because Social Security has not started yet — stays invested. Third, it shortens the horizon the portfolio has to fund, from 30 years of retirement to 29. With zero returns and zero inflation the first two alone put the retire-at-66 arm $90,000 ahead of the baseline at the end of age 66 in persona A ($97,000 in persona B), a figure the generator checks against the engine before publication; five years of saving $10,000 more puts the save-more arm $50,000 ahead at 65. Compounding then works on both, but the delay’s head start is larger and the horizon it has to carry is shorter.

The success rate shows the combined effect. In persona A, one year of work was worth +1.3 pp, 1.5× the +0.9 pp of a $10,000-a-year saving step; a second year added +0.9 pp more, and a second saving step +0.5 pp. In persona B the year of work was worth +11.5 pp against +6.7 pp for the saving step (1.7×), with the second year adding +7.1 pp and the second saving step +5.4 pp. At the median the year of work ended +$264,332 ahead at 95 in persona A and +$263,252 in persona B, against +$164,047 and +$155,645 for the saving step; in the bad tail (the 10th percentile of the per-path difference) the year of work was worth +$129,792 and $0, the saving step +$80,995 and $0.

Where spending cuts stand

A spending cut is the mirror image of a saving step: nothing at 65, then a little every year for life. $5,000 a year less is $150,000 of withdrawals avoided over 30 years in today’s dollars, more than the $50,000 of a saving step or the $90,000 first-year effect of a delay, but it arrives late, when much of it can no longer compound and, on the paths that fail, after the damage is done. In persona A the cut added +2.0 pp of success per $5,000 (+2.7 pp for $10,000), more than the saving step’s +0.9 pp and more than the year of work’s +1.3 pp. In persona B it added +12.9 pp (+21.9 pp for $10,000), more than the saving step’s +6.7 pp and more than the year of work’s +11.5 pp. Where the cut stands out is the median: at 95 it ended +$274,469 ahead in persona A and +$250,339 in persona B, the largest single-unit median gain of the three levers in one persona, because a cut that runs for 30 years is worth most on the paths that last the full 30 years.

The marginal plan

Every lever is worth more when the plan is close to the line. Persona B’s baseline succeeds in 70.9% of paths, so 2,912 paths are failing and many of them by a little; persona A’s 97.3% leaves only 272 to rescue, most of them badly. That is why the same year of work rescued 1,150 paths in persona B and 134 in persona A, and why the per-unit column is 8.6× larger for work, 7.5× for saving and 6.4× for spending in B than in A. The ranking of the levers per unit, spend less > work longer > save more, is the same in both; what changes is how much each is worth. The 10th-percentile wealth at 95 is $0 in every persona-B arm whose success rate is below 90% (more than one path in ten still fails), so the count of rescued paths is the better tail measure: 1,150 of 10,000 for one year of work, 1,857 for two, 665 and 1,201 for the saving steps, 1,292 and 2,187 for the spending cuts.

What it means

• A year of work is three levers in one. It adds a contribution, cancels a withdrawal and shortens the horizon. In these runs that made it worth 1.5× (persona A) and 1.7× (persona B) the success gain of saving $10,000 a year more for five years, and 0.7× and 0.9× the gain of spending $5,000 a year less.
• The units are not interchangeable dollars. A saving step is $50,000 before returns; a year of delay is $90,000 in its first year (persona A) plus a year less to fund; a $5,000 cut is $150,000 spread over 30 years. The per-unit column compares what each lever does, not what it costs the person — a year of work is a year of life at work, which the numbers do not price.
• The smallest-looking lever did the most per unit. $5,000 a year less spent moved success more than a year of work in persona A (+2.0 pp against +1.3 pp) and more than in persona B (+12.9 pp against +11.5 pp), because it runs for all 30 years of retirement ($150,000 undiscounted) and is worth most on the long paths; it rescued 203 / 1,292 failing paths against 134 / 1,150 for the year of work. Whether $5,000 a year for life is easier to give up than a year at work is a question the numbers do not answer.
• The marginal plan is where the levers matter. The same year of work was worth +11.5 pp in persona B and +1.3 pp in persona A; a comfortable plan has little left to gain from any of the three.
• Diminishing returns are mild over two units. The second year of work added +0.9 pp in persona A after the first’s +1.3 pp, and +7.1 pp after +11.5 pp in persona B; the second saving and spending steps behave the same way.
• Social Security was held still. Working to 66 or 67 did not change the benefit here. Delaying the claim as well would add the delayed-retirement credits on top — a separate lever, measured in the site’s claiming study.

CC-BY-4.0 — free for any use including republication and journalism, with attribution to QuantCalc Research.

Methodology

Assumptions and limitations

• The benefit is held at $30,000 from 67 in every arm. Working longer often goes with claiming later and with more earnings in the benefit formula; neither is modelled, so the work arms understate what a real extra year can do. The claiming-age lever is measured separately in the Social Security claiming study.
• Extra saving is affordable by assumption. The save-more arms add $10,000 or $20,000 a year without changing anything else; where that money would otherwise have been spent, the comparison with the work arms is between different sacrifices, which the numbers do not weigh.
• Constant-real spending. Every arm spends its planned amount every year with no response to the portfolio; a spending rule that adapts would change all six arms. The engine’s dynamic rules are compared in the withdrawal-strategy study.
• Parametric returns. Returns are lognormal with the stated means, volatilities and correlations; no regime switching, fat tails or historical block resampling. The sequence-of-returns study shows what resampled history does to a retirement date.
• Two personas, one mix. A 60/40 saver at 60 with five working years left. A different mix, a longer accumulation phase or a much larger balance would change the sizes, though not the mechanism.
• Pre-tax, no fees. No taxes, account types, fund expenses or trading costs on any arm.
• Not advice. Educational research on three ways of strengthening a retirement plan; not financial, tax or legal advice.

Frequently asked questions

Does working one more year or saving more move retirement success more?

Per unit, the order was the same in both personas: $5,000 a year less spent moved success the most, then a year of work, then $10,000 a year more saved for five years. Working longer beat saving more in both. For a 60-year-old with $800,000 on a 60/40 plan, saving $30,000 a year, spending $60,000 a year with $30,000 of Social Security from 67, retiring at 65 succeeded to 95 in 97.3% of 10,000 simulated paths. Retiring at 66 instead lifted that to 98.6% (+1.3 pp for one year), saving $10,000 a year more for the five working years lifted it to 98.2% (+0.9 pp), and spending $5,000 a year less lifted it to 99.3% (+2.0 pp). In the marginal persona ($600,000, spending $67,000, baseline 70.9%) the same three moves gave +11.5 pp, +6.7 pp and +12.9 pp. On the same market paths a year of work ended $264,332 higher at 95 at the median in persona A, against $164,047 for the extra saving and $274,469 for the spending cut.

Why does one more year of work do so much more than the same effort saved?

Because a year of delay does three things at once. It adds a year of contributions ($30,000), it removes a year of withdrawals ($60,000 at 65, before Social Security starts), and it shortens the retirement the portfolio has to fund from 30 years to 29. With zero returns the first two alone are worth $90,000 at age 66 in persona A, which the generator verifies against the engine; five years of saving $10,000 more is $50,000 at 65. The third effect, the shorter horizon, has no dollar figure but shows up in the success rate: in persona A one year of work was worth 1.5x the success gain of a $10,000-a-year saving step, and in persona B 1.7x.

How does spending less compare with working longer?

A $5,000-a-year cut is small in any one year but runs for the whole retirement: $150,000 of withdrawals avoided over 30 years in today's dollars, back-loaded rather than front-loaded. In persona A it added +2.0 pp of success (+2.7 pp for $10,000) against +1.3 pp for one year of work, so per unit a year of work was worth 0.7x a $5,000 cut. In persona B, where the plan is marginal, the cut added +12.9 pp (+21.9 pp for $10,000) against +11.5 pp for a year of work, a ratio of 0.9x. At the median the cut ended $274,469 higher at 95 in persona A and $250,339 in persona B, against $264,332 and $263,252 for the year of work.

Does the answer change when the plan is marginal?

The ranking held (spend less > work longer > save more per unit in persona B, spend less > work longer > save more in persona A), but every lever moved success far more when the baseline was 70.9% than when it was 97.3%: one year of work was worth +11.5 pp in persona B against +1.3 pp in persona A, $10,000 a year more saving +6.7 pp against +0.9 pp, and $5,000 a year less spending +12.9 pp against +2.0 pp. A comfortable plan has few failing paths left to rescue; a marginal plan has many near the line. The 10th-percentile wealth at 95 is $0 in every persona-B arm whose success rate is below 90% (more than one path in ten still fails), so the count of rescued paths is the better tail measure: 1,150 of 10,000 for one year of work, 1,857 for two, 665 and 1,201 for the saving steps, 1,292 and 2,187 for the spending cuts.

Does retiring later change the Social Security benefit in these runs?

No. Social Security is $30,000 a year from 67 in every arm, so the retirement date moves only when contributions stop and withdrawals start. Working to 66 or 67 and claiming at 67 are separate decisions here; a later claiming age would raise the benefit by the delayed-retirement credits and is the subject of the site's Social Security claiming study. Holding the benefit still isolates the work lever itself.

Are these numbers real dollars, and are taxes included?

All dollar figures are in today's dollars: the engine raises contributions, spending and Social Security with 2.5% inflation and deflates every reported balance back to today. Nothing is after tax, there is no pension, and every arm spends its constant-real amount every year; the study isolates the three levers and nothing else. Returns follow the JPMorgan LTCMA 2026 capital-market assumptions the site's planner uses, sampled with quasi-Monte Carlo, 10,000 paths per arm, identical draws for every arm of a persona.

Related research

Changelog

• v2026.1 (2026-09-08) — initial release. Two personas, six arms each (retire at 66/67, save $10,000/$20,000 a year more, spend $5,000/$10,000 a year less) against a retire-at-65 baseline; 10,000 paired paths per arm; success gained per unit; per-path paired statistics at 95 and by year.

Last updated 2026-09-08. Dataset license: CC-BY-4.0. QuantCalc is an independent retirement-planning research project. The JPMorgan LTCMA 2026 name identifies the published capital-market assumptions the engine uses; QuantCalc is not affiliated with, endorsed by, or sponsored by that firm, and all trademarks belong to their respective owners. Educational research, not financial, tax, or legal advice.

@misc{quantcalc2026workonemoreyearorsavemorewhichmovesretir,
  title  = {Work One More Year or Save More: Which Moves Retirement Success More? (2026)},
  author = {{QuantCalc Research}},
  year   = {2026},
  url    = {https://quantcalc.app/research/work-one-more-year-vs-save-more-2026/},
  note   = {Accessed <date>}
}