ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
Prolific Poll (US), follow-up survey (October 28 – November 2, 2026): full results
2
Accompanies our article "Election day poll: Have US voters changed their minds since last week?". We went back to the people who took part in our October 21–22 poll and asked them again. Every estimate reported in the article and its charts is in sheets A1–A10, with its 95% interval and unweighted base; the numbered sheets give weighted toplines and crosstabs, by party identification and 2024 presidential vote, for every closed question in the follow-up. The full questionnaire is published alongside this workbook; the October poll has its own workbook and questionnaire.
3
Survey
4
The October poll surveyed 1,518 US adult citizens (analysed) on October 21–22, 2026, through Prolific's opt-in participant panel, using propensity-score sampling. Between October 28 and November 2, 2026 we re-invited the 1,398 of them who had not told us they had already voted; 1,131 answered (80.9%). The 120 who had already voted by October were not re-invited and keep their October answers and weights in the estimates. The follow-up was a short online survey (median completion time about 1 minute): likelihood to vote, House vote, certainty, registration, Trump approval and the most important issue, plus an open question for people whose House vote or turnout answer had changed.
5
Data quality: completed interviews from test accounts were excluded; for the 2 people who completed the follow-up twice, the first completion is used; 3 completed interviews from people who were not part of the October analysis (excluded there as not US citizens) are not used. Topic-choice experiment (rows in sheets A3, A4 and A9): in October a random half of respondents were first asked which topic they wanted to answer questions about; see the October workbook.
6
Weighting
7
Each follow-up respondent starts from their October weight (raking to age × sex, race × education, region, employment, household income, 2024 and 2020 presidential vote, registration and 2026 party identification with leaners; see the October workbook). These weights are then adjusted for non-response: respondents are raked so that they match the October totals of everyone re-invited on the original weighting variables, on October House vote by October likelihood to vote, and on October House vote by registration; the adjustment is capped at 5 times the original weight. Weighted, the follow-up respondents represent US adult citizens who had not voted by October 22; together with the 120 early voters they represent all US adult citizens.
8
Populations
9
Follow-up respondents: the 1,131 people who answered the follow-up, weighted for non-response. Registered voters: follow-up respondents who say they are registered to vote at their current address (people who confirmed they had voted count as registered), plus the October early voters who said they were registered. Likely voters (headline): every person is weighted by their modelled probability of voting, from a model of validated turnout in the 2014, 2018 and 2022 Cooperative Election Study (past vote, registration, age, education, race, sex and stated intention to vote, with "I have already voted" as its own category), with turnout by age and education matched to the Census Bureau's 2022 pattern corrected to actual turnout (Hur and Achen) and overall turnout fixed at 47% of adult citizens.
10
Counting answers
11
Generic ballot: respondents who answered "Not sure" were asked which party's candidate they lean toward and are counted with that party. People who had already voted told us how they voted. D − R is the Democratic share minus the Republican share, in percentage points. Changes compare each person's follow-up answer with their own October answer.
12
Margins of error
13
Our sample is not a probability sample and is weighted, so the textbook margin of error does not apply. Following Sturgis and Kuha (Journal of the Royal Statistical Society, 2017), we estimated sampling variability by bootstrapping: October respondents were resampled with replacement within the 25 sampling groups 500 times, and the whole estimation (October weighting, the non-response adjustment and the turnout model) was repeated on each resample. ± is 1.96 times the bootstrap standard error (95% confidence); "95% CI" is the 2.5th to 97.5th percentile of the bootstrap estimates. For the dynamic MRP, intervals are 95% posterior intervals. These margins cover sampling and weighting variability only: not the choice of adjustment or model (alternatives in sheet A6), not changes among people who did not answer again that our data cannot see, and not reporting errors in either survey.
14
Bases
15
Unweighted n: number of people in the base. For likely voters, every person contributes in proportion to their probability of voting, so the unweighted n is everyone with a non-zero probability. Weighted n: sum of the weights in the base. Effective n: the sample size a simple random sample would need for the same precision, given the variation in weights. In sheets A2–A9, groups with fewer than 30 people are not reported; in the crosstabs, columns with fewer than 50 respondents are shown in grey italics. Participant counts are unweighted; weighted percentages represent the population named.
16
Open-ended answers
17
People whose House vote or turnout answer had changed were asked, in their own words, what had changed. 221 explanations from 196 people were coded by theme by an AI assistant (Claude) using a codebook of reasons; an independent second AI model coded a random 20% and agreed on the primary reason for 39 of 44 answers. Counts in sheet A10 are unweighted. Individual answers are not published, to protect participants; quotes in the article are lightly edited for spelling.
18
Rounding
19
All numbers are shown to one decimal place (the article rounds for readability). Shares may not sum to 100 because of rounding.
20
Sources
21
All estimates are computed in R. Wording of every question is taken from the fielded questionnaire. Self-reported voting and turnout will be checked after the election, when we re-interview everyone.
22
Contact: oriol.jover@prolific.com
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100