[SS-04] The Earnings Test, Zeros in Your PIA, and What FIRE Actually Costs

Retiring fifteen years early gives up 43% of the career and 28% of the benefit. The earnings test isn’t a tax. Both facts run the opposite way to the fear.

Pillar: Social Security & Medicare Enrollment · Applies to: Anyone with a short or interrupted career, anyone claiming before FRA while still working, and anyone weighing “one more year” Last verified: August 2026 · Refresh cadence: Annual — bend points and earnings-test thresholds via FN-02; the formula’s shape is statutory Related: SS-01 Claiming Age Math · SS-02 Spousal and Survivor Benefits · SS-03 Taxation of Benefits · ER-07 Coast and Barista FIRE · TR-03 One More Year · [IN-04] Part-Time Work · FN-02 Key Numbers

Not advice. The benefit formula uses indexed earnings, and the indexing stops at 60 — so the figures here, which use a flat career-average salary, illustrate the formula’s shape rather than predict your benefit. Pull your actual earnings record from SSA before relying on any of it, and check it for missing years while you still can.

---

TL;DR

• Two beliefs cost people real money, and both are backwards. That an early retirement guts your benefit, and that the earnings test confiscates it.
• The benefit formula is steeply progressive, so the years a FIRE retiree drops are credited at the lowest rate. Working 20 years instead of 35 at the same salary gives up 43% of the career and 28% of the benefit — $3,563 a month becomes $2,574.
• The marginal year is worth wildly different amounts depending on where you are: $91 a month after 20 years, $43 after 30, and exactly nothing after 35 at the same pay.
• The earnings test is a deferral, not a penalty. In §3, $17,760 is withheld in a year — and at FRA the benefit is recomputed upward for the 8.9 months withheld ✅. Almost nobody knows this, and people reshape their working lives around a confiscation that does not happen.
• Only your highest 35 years count, and earnings stop being wage-indexed at 60 ✅ — two reasons a late career year adds less than the salary suggests.

---

1. How the benefit is built

Three steps, and the shape of the second is the whole story.

1. Index your earnings to wage levels in the year you turn 60. Earnings after 60 count at nominal value ✅.
2. Average your highest 35 years, divided by 420 months ✅ — the AIME. Fewer than 35 years of earnings means zeros fill the gaps; the divisor never shrinks.
3. Apply the bend-point formula to get the PIA. For 2026 ✅ (FN-02): 90% of the first $1,286, 32% to $7,749, 15% above.

You need 40 credits — about ten years of work — to qualify at all ✅. Below that there is no benefit, and no partial one.

How credits accrue matters more than the headline suggests. One credit costs $1,890 of earnings in 2026 ✅ (FN-02), and a maximum of four count per calendar year ✅. So the full 40 requires $7,560 of earnings in each of ten years — a low bar in dollars and an unavoidable one in time. A high earner can buy all four credits in a single month and still cannot compress ten years into five. For anyone with a short or interrupted record, that makes the calendar the binding constraint, and it is the reason a modest part-time year is worth starting now rather than later.

2. What a short career actually costs

Career-average indexed earnings of $120,000, which puts this household in the top band.

Years worked	AIME	PIA	vs a full career
35	$10,000	$3,563	—
30	$8,571	$3,349	−6%
25	$7,143	$3,032	−15%
20	$5,714	$2,574	−28%
15	$4,286	$2,117	−41%

Giving up 43% of a career costs 28% of the benefit. The progressivity does it: a high earner’s final years are credited at 15 cents on the dollar, so those are the years dropped, and they were the cheapest ones to lose.

This is the number that should enter a FIRE decision, and the “fifteen zeros in your record” framing badly overstates it. The zeros are real, the divisor really is 420 months either way — and the formula still hands back a benefit far larger than proportional. Social Security is the most progressive asset a high earner owns, and early retirement is the moment that progressivity pays them.

The mirror image matters too. For a low earner the same table looks very different, because their marginal years land in the 90% or 32% band. Advice calibrated on a high earner’s benefit loss does not transfer downward, and this is one of the few places in the wiki where the answer flips by income.

3. The marginal year, which is not one number

One more year at the same salary, after…	Adds
20 years worked	$91/month
25 years worked	$91/month
30 years worked	$43/month
34 years worked	$43/month
35 years worked	$0/month

Three different answers to the same question. The 21st year is worth 2.1x the 31st, because the first still lands in the 32% band and the second has crossed into 15%. And the 36th year is worth nothing at the same pay — only the highest 35 count, so it replaces an identical year.

“One more year” is therefore a strong Social Security argument early in a career and an empty one at the end — the reverse of when it is usually deployed (TR-03). A 36th year only helps if it pays more than the lowest year currently in your 35, which for someone whose earnings peaked mid-career it often does not.

And the indexing cutoff compounds this. Earnings after 60 are counted at nominal value while every earlier year has been indexed up to age-60 wage levels ✅. A high-paying year at 63 is competing against inflated versions of your earlier years and frequently loses, which is why people who work to 65 sometimes find their statement barely moves.

4. The earnings test is not a tax

If you claim before FRA and keep working, benefits are withheld: $1 for every $2 above $24,480 in any year you are under FRA throughout ✅. In the year you reach FRA the threshold jumps to $65,160 and the rate falls to $1 per $3, counting only the months before your FRA month ✅. From FRA onward there is no test at all ✅.

Claiming at 63, benefit $2,000/month, still earning $60,000:

Earnings above the threshold	$35,520
Withheld at $1 per $2	$17,760
Equivalent to	8.9 months of benefit

Almost everyone believes that $17,760 is gone. It is not. At FRA, SSA recomputes the benefit and adjusts the early-claiming reduction to credit the months in which nothing was paid ✅ — producing a permanently higher monthly benefit from FRA onward.

So the earnings test is a deferral with a cash-flow cost, not a confiscation. The genuine costs are three: the money arrives later, it arrives as a slightly higher monthly benefit rather than a lump sum, and people reorganise their working lives to avoid a penalty that does not exist. That last one is the expensive part.

Two details that catch people. The test counts earned income only — wages and self-employment — so portfolio withdrawals, pensions, and Roth conversions are outside it entirely (ER-05). And it applies to spousal and survivor benefits drawn on your record too ✅ (SS-02), so a working claimant can suppress a spouse’s payment as well as their own.

5. The first year has its own test, and the self-employed test is hours

The annual threshold divided by twelve is $2,040 a month. In your grace year — the first year you are both entitled to a benefit and have at least one month under that limit — SSA can apply the test monthly instead of annually ✅. Any month you earn at or below $2,040 is paid in full, regardless of what you earned earlier in the year.

This is the rule that makes a mid-year retirement work at all. Earn $200,000 through June and retire: the annual test would wipe out every remaining cheque, while the monthly test pays July through December in full. Ask for it — it is not always applied automatically, and it applies once.

For the self-employed the test is hours, not dollars ✅: more than 45 hours a month in the business is substantial services, 15 to 45 hours is substantial in a highly skilled occupation, and under 15 hours never is. A consultant who bills almost nothing but stays involved can fail a test they assumed was about income.

Working after you claim still raises the benefit

SSA recomputes automatically each year. If a new year of earnings beats any of the 35 currently counted, it replaces the lowest, and the increase is effective January of the following year ✅.

Read that with §2 and the asymmetry appears again. A retiree with zeros in their record is replacing a zero — the most valuable substitution the formula permits. A 35-year career earner is replacing a real year and gains little or nothing. The same post-claim job is worth several times more to the short record, which is precisely the household most likely to think it is too late to matter.

6. What actually improves a short record

• Get to 40 credits. Below that the benefit is zero, not small ✅ — the single highest-value work anyone with a very short record can do.
• Fill zeros, not top years. A year that replaces a zero is worth multiples of one that replaces an existing year (§3). Part-time or consulting income in early retirement is worth more to your PIA than the same money earned at 34 years of service ([IN-04], ER-07).
• Check the earnings record for missing years. Employer reporting errors happen and are far easier to fix with contemporaneous evidence than decades later.
• Remember the spousal floor. A spouse with a thin record may do better on 50% of the higher earner’s PIA than on their own (SS-02) — in which case additional working years may add nothing at all to what they actually receive.
• Self-employment income counts, and so does the self-employment tax that comes with it ([IN-02]).

Three Scenarios

A. FIRE at 45 after 22 years at $120,000. She expects her benefit to be “destroyed.” It is about 28% below a 35-year career, arriving at 67 — a real reduction and a much smaller one than she budgeted for. The decision-relevant fact is that her dropped years were 15%-band years. Her plan improves more from delaying to 70 (SS-01) than from working three more years now.

B. Claims at 63, keeps consulting at $60,000. He is furious about $17,760 withheld and considers turning down work. The withholding is restored through a recomputation at FRA ✅, so the real question is only whether he wants the cash now or a higher benefit later — and the consulting income is also filling zeros and raising his PIA. Turning down the work would have been the costly move.

C. Barista FIRE at 52, twelve years of covered earnings. She is short of 40 credits and does not know it — below that threshold the benefit is not reduced, it is zero ✅. Two more years of even modest part-time earnings changes a nothing into a lifetime indexed income, and it is the highest-return work available to her (ER-07, [IN-04]).

💡 Pro-Tips

1. Pull your SSA earnings record now and check every year, especially early ones. Corrections get harder with time.
2. Count your credits before anything else if your record is short — 40 is a cliff, not a slope ✅.
3. Price a FIRE decision with the §2 table, not with the zero count. The progressivity is doing more work than the zeros are.
4. If your record is short, value part-time income as PIA repair, not just as cash ([IN-04]).
5. Do not turn down work to dodge the earnings test. It is a deferral and the benefit is recomputed at FRA ✅.
6. Only earned income counts for the test — withdrawals and conversions are outside it (ER-05).
7. Check whether the spousal benefit exceeds your own before valuing extra working years (SS-02).
8. Treat “one more year” as an income and healthcare argument, not a Social Security one, if you are past 35 years (TR-03).

⚠️ Common Pitfalls

1. Believing an early retirement guts the benefit. 43% of the career costs 28% of the benefit at this income (§2).
2. Believing the earnings test confiscates money. It defers it, and FRA restores it ✅.
3. Assuming a 36th year helps. At the same pay it adds exactly $0 (§3).
4. Forgetting the indexing stops at 60 ✅, so late high-earning years count at nominal value and often do not displace an indexed earlier year.
5. Applying high-earner conclusions to a low earner. The bands make the same table look completely different.
6. Not knowing you are short of 40 credits until you file ✅.
7. Claiming early while working and being surprised the spouse’s benefit is also suppressed ✅ (SS-02).
8. Never checking the earnings record and discovering a missing year when the employer no longer exists.
9. Treating the AIME divisor as flexible. It is 420 months regardless of how long you worked.
10. Ignoring that the trust-fund question sits underneath all of this (SS-01 §7, FN-03) — these are scheduled benefits, not guaranteed ones.

✅ Actionable Checklist

Once, this month

• Download your SSA earnings record and verify every year against your own records
• Count your credits; confirm you are at or above 40 ✅
• Note how many of your 35 computation years are currently zeros

If you are considering retiring early

• Compute the PIA at your actual career length and at the length you would reach by working longer (§2)
• Compare that difference against what delaying the claim to 70 buys (SS-01)
• Check whether the spousal benefit would exceed your own anyway (SS-02)

If you will claim before FRA and keep working

• Project earnings against the $24,480 threshold ✅
• Confirm which income counts — earned only (ER-05)
• Calendar the FRA recomputation so you can verify SSA applied it
• Check whether anyone claims on your record and would be suppressed (SS-02)

In early retirement with a short record

• Treat part-time or consulting income as filling zeros, and value it accordingly ([IN-04])
• Re-pull the earnings record every few years to confirm the new years posted

Sources & further reading (verified August 2026)

• SSA, “Your Retirement Benefit: How It’s Figured” and the Annual Statistical Supplement, Appendix: Computing a Retired-Worker Benefit — indexing of earnings to age-60 wage levels with later years at nominal value, the highest-35-years rule, the 420-month divisor, and the bend-point formula.
• Congressional Research Service, R41242, “Social Security Retirement Earnings Test: How Earnings Affect Benefits” — the $1-per-$2 and $1-per-$3 withholding rules, and the recomputation at full retirement age that adjusts the actuarial reduction for months in which benefits were withheld.
• Congressional Research Service, R46658, “Social Security: Benefit Calculation” — the progressive structure of the PIA formula and its distributional effect.
• SSA, “How You Earn Credits” — the 40-credit requirement for retirement benefits.
• FN-02 for the 2026 bend points ($1,286 / $7,749) and earnings-test thresholds ($24,480 / $65,160).
• tools/ss04_worked_examples.py — every figure in §2–§4, computed and checked against this article by CI. The salary is a stated input; the bend points and thresholds come from FN-02.
• SS-01 owns the claiming-age decision · SS-02 owns spousal and survivor entitlement, including the floor that can make extra working years irrelevant · ER-07 owns the Coast/Barista decision this page supplies the benefit arithmetic for · TR-03 owns “one more year” as a whole.