ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
Prolific Poll (UK) — September 2026
2
Political donations: full data tables
3
4
Commissioned byNone. This survey was designed, funded, conducted and published by Prolific for its own research. There is no external client or sponsor.
5
Conducted byProlific
6
Fieldwork21 September 2026
7
Method of interviewOnline, self-administered; survey scripted in Qualtrics
8
Universe representedUK adults aged 18 and over
9
Sample size1,500 UK adults. Each respondent evaluated six hypothetical donations, giving 9,000 vignette evaluations in total.
10
SamplingProlific opt-in online participant panel, recruited to quotas for age, sex and region. Age and sex come from Prolific's participant-level profile data and were not re-asked in this survey.
11
WeightingWeighted to the UK adult population on age and sex, region, education by age, housing tenure, and 2024 general election vote, using regularised raking. Effective sample size approximately 912; design effect 1.64; maximum weight 4.75, with no respondent weighted above five times the average.
12
Weighted or unweightedThe experimental results in this workbook are UNWEIGHTED. Treatment was randomised within each respondent, so the comparisons are identified by that randomisation rather than by the composition of the sample. Weighted equivalents are reported alongside as a sensitivity check wherever they were produced. The two policy questions are WEIGHTED, because there the aim is to describe the population rather than to estimate a randomised effect. Every table states which applies.
13
Margin of errorAs with any sample survey these results are subject to a margin of error. For a finding of 50% on the full weighted sample this is approximately +/-3.2 percentage points at the 95% confidence level; findings further from 50% carry a slightly smaller margin of error. As the British Polling Council notes, all polls are subject to a wide range of potential sources of error beyond sampling variation alone.
14
Questions coveredThe six-vignette donations experiment (DON_V1 to DON_V6), the perceived-influence question (DON_INFLUENCE), and the two policy questions (DON_CAP_SUPPORT, DON_PUBLIC_FUNDING). The full questionnaire is published separately.
15
Published alongside“Britons don't think large donations from overseas donors should be legal. Except Reform voters”, Prolific, September 2026.
16
BPC membershipProlific is a registered member of the British Polling Council and abides by its Rules of Disclosure. This workbook contains the full response distributions, weighted and unweighted bases, demographic crosstabs and model estimates for every question covered. Anything further, including a fuller description of the weighting procedure, is available on request.
17
Contactoriol.jover@prolific.com
18
19
Contents
20
21
DesignThe three randomised elements, their levels, and the within-respondent balance
22
Exp_GridTable 1. Acceptance by amount × donor profile, the 18-cell grid
23
Exp_AmountTable 2. Acceptance by amount, raw and model-adjusted, plus contrasts
24
Exp_DonorProfileTable 3. Acceptance by donor profile, plus residence and registration contrasts
25
Exp_RecipientPartyTable 4. Acceptance by recipient party and own-party effects
26
Exp_Amount_byVITable 5. Amount by current voting intention, with per-tenfold slopes
27
Exp_Donor_byVITable 6. Donor profile by current voting intention, with effects
28
Model_estimatesTable 7. All model contrasts and specification tests (LEGAL outcome)
29
Policy_ToplinesTable 8. Headline support for a cap and for public funding
30
Dist_LEGALTable 9. Full four-point distribution, primary outcome
31
Dist_ACCEPTTable 10. Full four-point distribution, secondary outcome
32
Exp_Grid_byPartyTable 11. All 36 experimental combinations
33
DON_INFLUENCETable 12. Perceived influence of large donors, 0-10
34
Policy_DistributionsTable 13. Policy questions in full, and the question-order effect
35
Policy_DemographicsTable 14. Policy questions by demographic group
36
Exp_DemographicsTable 15. Experiment results by demographic group
37
Model_estimates_ACCEPTTable 16. Model estimates for the secondary outcome
38
39
How to read this workbook
40
41
The experimentEvery respondent judged six hypothetical donations. Each one is a single sentence built from three independently randomised elements: the amount, the donor's connection to the UK, and the recipient party. The six were balanced within each respondent, so everyone saw each amount once, met each donor profile twice, and judged three donations to Labour and three to Reform UK. See the Design sheet.
42
Acceptance ratesUnless a table says otherwise, the figure shown is the share saying a party should be “legally allowed” to accept the donation, that is the two categories “Definitely should allow” and “Probably should allow” combined. The underlying question used a four-point scale.
43
Shares and effectsTwo different quantities appear in this workbook, and they are labelled separately. A share is a percentage of people or of answers. An effect is a difference between two shares and is reported in percentage points, with a 95% confidence interval.
44
BasesTables built on the grid of amount by donor profile carry at most one answer per respondent per cell, so the respondent base and the answer base are the same. Tables pooled across amounts carry two answers per respondent per donor profile; those state an answer base. Both are given.
45
Model estimatesWithin-person estimates come from linear probability models with respondent fixed effects and standard errors clustered by respondent. See the Model estimates sheet.
46
RoundingPercentages are rounded to one decimal place. Rounding can produce totals that do not sum to exactly 100%.
47
CautionThis is an opt-in online panel, not a random probability sample. The strongest claims in this workbook are the randomised within-person comparisons. Absolute levels, and generalisation to the wider population, warrant more caution. Donations described in the experiment were hypothetical.
48
Subgroup scopeTwo deliberate choices. The policy-question demographic breakdowns (Table 14) pool both question orders, because splitting by order as well would leave several cells too thin to report. The experiment demographic crosstabs (Table 15) exclude current voting intention, which has its own fuller treatment in Tables 5 and 6.
49
Empty cellsWhere no respondent in a group selected a given answer the figure is shown as 0.0% and no confidence interval is reported, because none can be estimated. Four cells on the DON_INFLUENCE sheet are affected and are listed there.
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