Hawaiʻi's 2026 Primary Ballot Curve, Day by Day: The Benchmark to Measure Your Chase Against

Every campaign watches its own ballot return numbers. Far fewer know what those numbers should look like on a given day.

We archive a snapshot of the official Hawaiʻi Office of Elections absentee-reconciliation reports twice a day. That archive now holds 204 distinct report dates going back to March 31, 2026 — 24 of them covering the closing stretch of the August 8 primary cycle, from D-23 to D+7, listed in full below. That is enough to answer the question every campaign is quietly asking: am I ahead or behind?

Historical data. Every number in this post is a record of what the official reconciliation reports actually showed on the dates listed below. None of it moves. It is here as a benchmark for campaigns watching the 2026 general cycle, which is the current live cycle — for current numbers, see Hawaiʻi 2026 Ballot Returns by Precinct.

The primary curve, day by day

Election day for the 2026 primary was August 8. Here is the real statewide curve. The early rows matter: they show the long flat stretch, which is the part every campaign misreads as “nothing is happening.”

Days out	Date	Sent	Returned	Return rate	Invalid
D-23	Jul 16	728,224	31	0.00%	0
D-19	Jul 20	728,546	94	0.01%	0
D-18	Jul 21	729,525	2,869	0.39%	27
D-17	Jul 22	730,159	20,554	2.82%	175
D-16	Jul 23	730,561	42,162	5.77%	456
D-15	Jul 24	731,108	54,978	7.52%	584
D-12	Jul 27	732,320	75,630	10.33%	804
D-11	Jul 28	732,862	87,510	11.94%	908
D-10	Jul 29	733,460	101,341	13.82%	1,014
D-9	Jul 30	734,045	115,132	15.68%	1,124
D-8	Jul 31	734,533	126,849	17.27%	1,209
D-7	Aug 1	735,154	136,926	18.63%	1,287
D-5	Aug 3	735,653	145,485	19.78%	1,352
D-4	Aug 4	736,523	159,276	21.63%	1,416
D-3	Aug 5	737,119	179,223	24.31%	1,543
D-2	Aug 6	737,663	198,855	26.96%	1,710
D-1	Aug 7	738,334	214,788	29.09%	1,812
D0	Aug 8	739,372	234,639	31.73%	1,929
D+2	Aug 10	741,689	280,477	37.82%	2,565
D+3	Aug 11	741,690	280,567	37.83%	2,499
D+4	Aug 12	741,690	280,736	37.85%	2,435
D+5	Aug 13	741,690	280,923	37.88%	2,248
D+6	Aug 14	741,690	281,079	37.90%	2,092
D+7	Aug 15	741,690	281,249	37.92%	1,922

Rows in bold are the inflection: the primary went from 0.39% to 7.52% in three days, between D-18 and D-15. Everything before that was flat, and everything after was a steady climb.

Four things in that table matter more than the numbers themselves.

1. More than half of all ballots arrived in the final stretch

At D-8, 126,849 ballots had come back. By D+7, 281,249 had. The difference — 154,400 ballots — is 54.9% of the entire primary’s mail-ballot return, delivered across the fifteen days from July 31 to August 15.

This is the single most useful thing in the table. If your campaign treats the first three weeks of the cycle as the “main” return period, you have your staffing and your contact schedule backwards. The first three weeks are when you get ballots into hands. The final stretch is when you get them back.

2. Election day is not the peak — the days after it are

Here is where campaigns misread their own data. Pulling the raw increment for election day gives 19,851 ballots in a single day. It is tempting to call that the peak. It is not.

The next report is two days later (the Office did not publish an Aug 9 report), and it adds 45,838 ballots across those two days — an average of 22,919 per day, the highest sustained daily pace of the entire cycle.

The cleanest way to see it is to compare equal three-day windows on either side of election day:

Window	Days	Ballots returned	Per day
D-8 → D-5	3	18,636	6,212
D-1 → D+2	3	65,689	21,896

The post-election window ran at 3.5x the pace of the comparable pre-election window. About one ballot in six of the cycle’s entire mail-ballot return arrived after the polls closed — 46,610 of 281,249, or 16.6%, spread thinly across D+1 through D+7.

The operational consequence: staff your signature-verification and ballot-processing surge for the days after election day, not for election day. If your volunteers are finished by 7pm on Nov 3, you will be sitting on tens of thousands of mail ballots with nobody checking signatures.

A caution about reading these numbers: the Office does not publish a report every single day. Two of our windows span two days, and comparing a two-day increment against a one-day increment will make any curve look dramatic. Always normalize to a per-day rate before drawing a conclusion.

3. The curve is a cliff, not a slope

Normalized to ballots per day, the whole cycle:

D-7   +10,077/day   (1-day window)
D-5    +4,280/day   (2-day window)
D-4   +13,791/day   (1-day window)
D-3   +19,947/day   (1-day window)
D-2   +19,632/day   (1-day window)
D-1   +15,933/day   (1-day window)
D0    +19,851/day   (1-day window)
D+2   +22,919/day   (2-day window)  <-- sustained peak
D+3       +90/day
D+4      +169/day
D+5      +187/day
D+6      +156/day
D+7      +170/day

Returns rise roughly 4.7x between D-5 and D-3, hold around 20,000 a day through election, tick up slightly after it, then collapse to under 200 a day for the rest of the cycle.

Once you see this shape, “my returns are low today” stops being alarming and starts being expected. The days that should worry you are the ones where the daily rate falls to near zero while you are still inside the final week.

4. The cure process is visible in the data, and it works

The invalid-ballot column tells a story most campaigns never see. Invalid ballots climbed steadily to a peak of 2,565 on D+2 — a rejection rate of 0.91%. Then they went down, to 1,922 by D+7.

Nothing about returning ballots should make them valid. The only reason that number falls is that election officials cured them: they contacted voters, voters responded, and ballots that were on track to be rejected came back into the count.

643 ballots were rescued in five days. For a primary decided by margins in the dozens, that is not a rounding error.

Note also that the invalid rate was already falling before election day, from 0.95% at D-8 to 0.82% at D0 — the denominator was growing faster than the numerator, which is what a well-run mail-ballot program looks like.

Your benchmark for November 3

The general election is November 3. Apply the primary’s shape — but check the calendar, because the two cycles are not tracking each other.

Comparing both cycles at equal days-out, from the official reports:

Days out	Primary return rate	General return rate	Gap
D-19	0.01%	0.07%	+0.06 pp
D-18	0.39%	0.11%	−0.28 pp
D-17	2.82%	—	—
D-16	5.77%	—	—
D-15	7.52%	2.05%	−5.47 pp
D-14	no primary report	4.74%	—
D-13	no primary report	8.03%	—

Two things follow.

The shape still holds. The primary accelerated sharply from D-18 onward, hitting 7.52% by D-15. The general cycle has run that same acceleration: 0.11% at D-18, 2.05% at D-15, 4.74% at D-14, 8.03% at D-13. Expect the climb to continue on the primary’s pattern:

• Now through D-8: expect roughly 17% of mail ballots returned.
• D-8 to D0: expect the return rate to climb toward 32%.
• D+2: expect the highest sustained daily pace of the cycle, and staff for it.
• D+3 and after: returns collapse to near zero. Your ballot-chase operation should be winding down, and your signature-cure and audit operation should be winding up.

But the two cycles were not tracking each other, and have now converged. The gap at D-15 was real and it mattered: if you had been benchmarking your general-cycle numbers against the primary’s absolute percentages, you would have looked badly behind schedule while sitting exactly on the primary’s curve.

That gap has since closed. On D-13 the general cycle reached 8.03% — above the 7.52% the primary was sitting at on D-15. Two cycles that looked nothing alike four days ago now track closely. So the practical rule depends on when you are looking:

• Early in the cycle: benchmark against the primary’s rate at the same D-x, not its absolute numbers.
• From roughly D-13 onward: the two curves have converged, so the primary’s absolute percentages are finally usable as a direct target.

The convergence is the single most useful update for anyone using this post as a live benchmark.

Compare your own numbers to these, not to a national average. Hawaiʻi’s curve is Hawaiʻi’s curve.

What we cannot tell you

Our automated archive starts collecting on March 31, 2026, but the earliest primary-cycle reports it holds begin July 16 — D-23. So we have the primary’s curve from D-23 forward, including the entire inflection, but not the full front of the cycle.

What the archive cannot tell you is anything before D-23. If your campaign wants a multi-cycle historical baseline, we do not have one. And a single primary is a single sample: the shape below is a useful benchmark, not a law.

Check the live numbers

• Ballot returns by precinct — checked every 30 minutes; the page updates itself when the Office publishes a new report, and states its own report date so you can see how current it is
• Precinct map
• Office of Elections absentee reporting — the primary source, unmodified

This is a turnout-planning tool. Use it to find voters who need help participating, never to target persuasion against a demographic or a neighborhood.

Primary curve reconstructed from Office of Elections reconciliation snapshots (report dates 2026-07-16 through 2026-08-15 — 24 report dates in all). The 13 rows from 2026-08-01 onward are verified against our own archived data module; earlier rows were read directly from the Office’s published PDFs. Our automated archive holds 204 distinct report dates beginning 2026-03-31. Per-day rates normalize for multi-day reporting gaps in the sequence.