The final snapshot.
Question 10: remain or begin the referendum process
Decided respondents · question-specific unweighted bases not supplied
Undecided treatment is identified in the response-base and methodology records. Percentages may be rounded.
Source: Northbound Insights, supplied report figures. Response base: Decided respondents · question-specific unweighted bases not supplied. Subgroup interview counts were not supplied.
The large Question 10 gap is distinct from the much closer constitutional proposals. Each question has its own response base and should be read separately.
Ten questions.
Different decisions.
One electorate.
Different patterns.
Reported group differences are descriptive. Smaller bases, weighting and selection can increase uncertainty; no design-adjusted significance tests are claimed.
A three-to-one Question 10 estimate
The final 75–25 split concerns remaining in Canada versus beginning the constitutional process for a later binding referendum. Question 10 did not itself ask voters to leave Canada immediately. Its 50-point gap is much larger than the differences on the other supplied questions.
The autonomy questions do not have one uniform level of support
Proof of citizenship for provincial voting has the largest supplied yes share at 62%. The four immigration and services items range from 52% to 55%. The judicial-appointment, Senate, federal-program and provincial-law questions range from 49% to 53%, with much smaller point-estimate gaps.
Regional differences depend on which question is being asked
The supplied remain figures are 82% in Calgary, 85% in Edmonton and 61% in rural Alberta. On Questions 1–9, the rural yes estimates are higher than the two city estimates. A single general description of an autonomy voter would obscure those question-specific differences.
The August comparison is a limited additional observation
The supplied narrative gives an August remain baseline of 69%, six points below the final estimate. Its exact field dates, release date, sample size and recruitment details were not provided, so it is retained as context rather than plotted as a fully documented tracking wave. No individual voter movement is inferred.
Ten ballots require ten denominators
The supplied narrative proposes 20–30% ballot drop-off on technical questions, but it does not include the underlying turnout or ballot-completion model. The charts do not adjust results using that assumption. Each question’s undecided and blank-response counts would be needed to estimate an all-respondent ballot-completion pattern.
How to read
this study.
The supplied methodology is preserved alongside its disclosure gaps, so the total sample, ballot base and subgroup estimates can be assessed separately.
Study design and disclosure · n=1,100
- Research provider
- Northbound Insights
- Report provenance
- Figures, release label and methodology supplied by Northbound for this website update on 25 October 2026. The underlying interviews have not been independently audited here.
- Release date
- 18 Oct 2026
- First publication of this web page
- 25 Oct 2026
- Fieldwork
- 14 Oct 2026 – 17 Oct 2026
- Population
- Eligible provincial voters in Alberta
- Total interviews
- 1,100
- Ballot denominator
- Decided respondents · question-specific unweighted bases not supplied. Effective sample size was not supplied.
- Undecided
- Not supplied
- Interview mode
- 50% live-agent telephone interviews; 50% SMS-to-web panels.
- Recruitment and sampling
- Described as a stratified random sample of eligible Alberta voters. Telephone selection, SMS-panel recruitment and inclusion probabilities were not supplied.
- Weighting
- Age, gender, education and region: Calgary, Edmonton, Other Cities and Rural Alberta, using Elections Alberta elector distributions. Numerical targets and design effects were not supplied.
- Supplied precision statement
- ±3.0 percentage points, 19 times out of 20
- Uncertainty assessment
- The release reports ±3.0 pp, 19 times out of 20. A conventional margin is not assigned here to a mixed telephone and SMS-panel estimate without documentation of probability recruitment for all components. It cannot be transferred to smaller question-specific or regional bases.
- Commissioning / funding
- Not identified in the supplied material.
- Questionnaire / response rates
- Exact wording, order, response options, response and completion rates were not supplied.
- Subgroup bases / design effects
- Unweighted and weighted subgroup bases, weighting distributions and design-effect estimates were not supplied.
- Tracking design
- No earlier comparable wave was supplied for this report.
Sample size alone does not validate a probability margin for a mixed-mode study. Reported margins are retained as source statements in the disclosure record. They are not used as confidence intervals for the charts, point-estimate gaps or subgroup estimates.
The figures,
open to inspection.
- Northbound Insights · report figures and methodologyRelease 18 Oct 2026 · supplied for this website on 25 Oct 2026. Earlier-wave metadata is left blank where unavailable.
- Official election / referendum information ↗Election context and candidate records; not a source or endorsement of this poll.
This report presents the figures supplied by Northbound. It does not reconstruct respondent records, infer missing categories or claim independent verification of the fieldwork. A point-estimate lead is not a forecast of the outcome. Publication standards.