Original research
2026 Voter Contact Benchmarks
Across 1,210,664 dial attempts in six 2026 primary phone bank programs, 3.4% of dials reached a live voter who spoke with an agent, 56% went to voicemail, 33% of the voters who answered refused, and 24% of live conversations ended with a recorded voter position. A typical program (per program mean) answered live at 5.2%.
By Frank Carrasco, PhD, founder of The Political Group. Published . Data: April 4 to August 3, 2026. Free to cite with attribution (CC BY 4.0).
What happens to a political phone call
| Outcome | Pooled | Per program mean | Range |
|---|---|---|---|
| Live answer, connected to an agent | 3.4% | 5.2% | 2.6% to 9.9% |
| Voicemail or answering machine | 56.3% | 63.5% | 54.0% to 71.9% |
| No answer | 36.5% | 24.7% | 14.1% to 41.4% |
| Busy | 1.0% | 1.8% | 0.7% to 3.3% |
| Disconnected or bad number | 1.8% | 2.7% | 0.6% to 5.8% |
| Human picked up, no agent free (system drop) | 1.0% | 2.1% | 0.7% to 5.5% |
System drops are human pickups lost because no agent was free: 22% of all human pickups pooled. Including them, humans picked up 4.3% of dials. That gap is the cost of over-aggressive dialer pacing and is the first number to check on any vendor's dashboard.
What voters do once they answer
| Metric | Pooled | Per program mean | Programs |
|---|---|---|---|
| Refusal share of live answers (hang-up, not interested, do-not-call) | 33.0% | 26.7% | 5 |
| Hang-up share of live answers | 23.9% | 18.0% | 5 |
| Explicit not-interested share of live answers | 6.5% | 6.1% | 5 |
| Do-not-call request share of live answers | 2.6% | 2.6% | 5 |
| Callback request share of live answers | 22.3% | 28.3% | 5 |
| Voter ID yield: share of live answers that ended in a recorded position | 24.4% | 34.1% | 6 |
| Support share of live answers | 6.5% | 12.5% | 6 |
| Undecided share of live answers | 4.4% | 6.8% | 6 |
| Opposed share of live answers | 4.9% | 5.3% | 6 |
How identified voters split
| Metric | Pooled | Per program mean | Programs |
|---|---|---|---|
| Support share of identified voters | 26.7% | 32.7% | 6 |
| Lean / considering share of identified voters | 8.5% | 8.7% | 6 |
| Undecided share of identified voters | 18.2% | 19.9% | 6 |
| Opposed / not supporting share of identified voters | 20.1% | 20.1% | 6 |
| Not interested share of identified voters | 26.5% | 18.6% | 6 |
| Support share among decided voters | 57.0% | 63.5% | 6 |
| Voter IDs per dial attempt | 0.8% | 1.6% | 6 |
Live answer rate by hour
One statewide U.S. Senate program, 15 field days, hours with 500+ attempts. Hours are the dialer clock (Eastern); the electorate was two time zones west.
| Hour (ET) | Attempts | Live rate |
|---|---|---|
| 1 pm | 3,945 | 4.6% |
| 2 pm | 7,135 | 4.7% |
| 3 pm | 9,238 | 4.8% |
| 4 pm | 11,381 | 5.1% |
| 5 pm | 13,084 | 4.8% |
| 6 pm | 13,109 | 4.7% |
| 7 pm | 15,785 | 6.2% |
| 8 pm | 18,627 | 6.5% |
| 9 pm | 21,306 | 5.7% |
| 10 pm | 6,082 | 3.1% |
Live answer rate by weekday
Same program. Sunday had only 1,782 attempts and is listed for completeness.
| Day | Attempts | Live rate |
|---|---|---|
| Monday | 10,680 | 6.2% |
| Tuesday | 15,782 | 5.1% |
| Wednesday | 24,747 | 6.0% |
| Thursday | 22,432 | 4.2% |
| Friday | 25,802 | 5.6% |
| Saturday | 18,890 | 5.5% |
| Sunday | 1,782 | 1.4% |
Average live answer lasted 40 sec; a conversation that produced a voter ID lasted 78 sec (two call log programs).
The programs behind the numbers
| Program | Window | Field days | Attempts | Live answers | Live rate |
|---|---|---|---|---|---|
| US House primary | 2026-05-21 to 2026-05-30 | 10 | 824,691 | 21,319 | 2.6% |
| US House primary | 2026-06-01 to 2026-06-17 | 6 | 204,705 | 8,267 | 4.0% |
| US Senate primary | 2026-04-04 to 2026-04-26 | 15 | 120,115 | 6,382 | 5.3% |
| US House primary | 2026-08-02 to 2026-08-03 | 1 | 34,958 | 3,476 | 9.9% |
| County commissioner primary | 2026-06-01 to 2026-06-04 | 2 | 20,649 | 1,289 | 6.2% |
| County council primary | 2026-04-25 to 2026-04-26 | 2 | 5,546 | 183 | 3.3% |
All six were agent staffed, predictive dialed programs calling voter file phone numbers on behalf of a candidate in a 2026 primary. Candidates are not named; the data is theirs. Pooled figures are volume weighted and the largest program is 68% of attempts, so the per program mean is the fairer picture of a typical program.
Methodology
Source. The dialer's own disposition code on every outbound attempt: per call log exports for two programs, campaign wide status counts from the dialer API for the other four. Nothing was sampled, modeled, or filled in.
Live answer. A human answered and was connected to an agent, whatever they then said. Voicemail, no answer, busy, disconnected numbers, and human pickups that the dialer dropped for lack of a free agent are not live answers.
Identified voter. A live answer that ended with a recorded position: support, lean, undecided, opposed, or explicitly not interested.
Aggregation. Each rate metric reports a pooled value (sum of numerators over sum of denominators, weighted by volume) and a simple_mean (unweighted average of per-campaign rates). The pooled figure is dominated by the largest campaign (824,691 attempts); the simple mean treats each program equally.
Not computed. 7 metrics were left out because the data does not support them, including a "not voting" share (no such disposition exists) and per number reach for the API snapshot programs. The full list is in the dataset file.
Cite as. The Political Group, "2026 Voter Contact Benchmarks," September 2026, https://www.thepoliticalgroup.com/research/2026-voter-contact-benchmarks. Licensed CC BY 4.0.
Questions this data answers
What is a typical live answer rate for political phone banking?+
Across six 2026 primary phone programs and 1,210,664 dial attempts, 3.4% of dials reached a live person who spoke to an agent (volume weighted). The per program average was 5.2%, with a range of 2.6% to 9.9%. Voicemail took 56% of dials and no answer 37%.
How many phone calls does it take to identify one voter?+
Voter IDs per dial attempt were 0.82% pooled and 1.57% per program, which is roughly one recorded voter position for every 122 dials pooled or every 64 dials for a typical program. Of the people who answered live, 24% ended the call with a recorded position.
What share of voters refuse a political call?+
33% of live answers ended in a hang up, an explicit not interested, or a do not call request (pooled across five programs). Hang ups alone were 24% of live answers, explicit do not call requests 2.6%. A further 22% asked for a callback.
What time of day do voters answer political calls?+
In the one statewide program with hour level data, the best live answer rate was 6.5% at 8 pm on the dialer clock (Eastern), against 4.6% at 1 pm. Evening hours from 7 pm onward outperformed every afternoon hour. Monday had the best weekday rate at 6.2% and Thursday the worst at 4.2%. This is a single campaign and should be treated as indicative.
How were these benchmarks computed?+
From the dialer's own status codes on every dial in six agent staffed, predictive dialed 2026 primary programs run by The Political Group (three U.S. House, one U.S. Senate, one county commissioner, one county council). Live answer means a human answered and was connected to an agent. Pooled figures divide total numerators by total denominators, so the largest program carries most of the weight; the per program mean treats each program equally. No estimates, no modeling, no demo data.
Plan a program against these numbers
Size the dials and callers you need, then price the list.