# EXECUTIVE POLLING INTELLIGENCE BRIEF
## 2021 Alberta Equalization Referendum: Polling vs. Actual Results Audit

**Course:** LIAI 1001 – Strategy and Change  
**Author:** Graham Turnbull (Student ID: 200679442)  
**System:** AI Polling Intelligence Agentic System (PollBot Analyst)  
**Subject:** Viewpoint Alberta Survey (Leger / U of A) vs. Certified Referendum Results  

---

### 1. Executive Summary & Comparative Matrix

On October 18, 2021, Albertans voted on removing the principle of equalization payments from the Canadian Constitution. Pre-referendum polling reported in the *Viewpoint Alberta Research Brief* (Page 2, Figure 1) achieved **exceptional statistical accuracy among decided voters**, but presented an **analytically flawed narrative** that framed the electorate as closely divided.

| Metric | Viewpoint Alberta Poll<br>*(All Adults Sample)* | Viewpoint Alberta Poll<br>*(Decided Voters Recomputed)* | Official Certified Results<br>*(October 18, 2021)* | Statistical Divergence |
| :--- | :---: | :---: | :---: | :---: |
| **"Yes" Vote** | 43.0% | **62.3%** | **61.7%** | **+0.62 pp** |
| **"No" Vote** | 26.0% | **37.7%** | **38.3%** | **-0.62 pp** |
| **Lead / Spread** | +17.0 pp | **+24.6 pp** | **+23.4 pp** | **+1.24 pp** |
| **Undecided / Non-Vote** | 28% DK + 3% Won't Vote (31.0%) | — *(Filtered)* | 4.52% *(Blank/Declined)* | — |
| **Base / Turnout** | $N = 1,204$ (Weighted) | $n = 831$ (Decided) | 1,092,704 ballots cast | — |

> [!NOTE]
> **Key Benchmark:** With decided voters isolated ($\frac{43}{69} = 62.3\%$), the poll predicted the final outcome within **0.6 percentage points**, well inside the survey's theoretical margin of error ($\pm 2.8\%$).

---

### 2. What the Poll Got Right

1. **Pinpoint Decided-Voter Accuracy:**  
   The underlying sample accurately captured the true $62/38$ proportional preference of the electorate.
2. **Key Demographic Cleavages:**  
   The survey correctly predicted the core pillars of the "Yes" coalition:
   - **Age Gradient:** Senior voters (65+) showed the strongest support at **58% Yes**.
   - **Regional Core:** Central (**57% Yes**) and Rural Alberta (**48% Yes**) were the leading geographic bases.
   - **Gender Gap:** Identified a 13-point gender split (**51% Men Yes** vs. **38% Women Yes**).
3. **Turnout Salience Advantage:**  
   The poll detected that "Yes" supporters placed a much higher personal importance on the vote (**8.1/10**) than "No" supporters (**4.0/10**). When filtered by voters *"certain to vote"* (Page 6), the decided ratio stood at **62.8% Yes**, matching the actual outcome.

---

### 3. What the Poll & Commentary Got Wrong

1. **The "Divided Electorate" Narrative Fallacy:**  
   The academic brief was titled *"Albertans & Equalization: Divided and Misguided?"* and argued that public opinion was deeply divided based on raw figures (43% Yes vs. 26% No). In reality, the referendum was a **landslide 23.4-point blowout** (642,501 to 399,169 votes). Emphasizing the raw 43% obscured what was already a +24-point decided supermajority.
2. **Failure of the Traditional "Status Quo" Rule:**  
   Referendum lore assumes undecided voters break toward the status quo ("No") out of risk aversion. Here, undecideds either stayed home, cast blank ballots, or broke in line with decided voters.
3. **Under-Emphasizing the Likely Voter Screen:**  
   Leading with the broad adult population rather than the "certain voter" screen masked the full momentum of the mobilized electorate.

---

### 4. Non-Sampling Ballot Artifacts

* **Elevated Blank & Declined Ballots (49,401 Ballots / 4.52%):**  
  Unlike typical elections where spoiled/blank ballots are $<1\%$, over 49,000 Albertans declined or skipped the referendum ballot, reflecting voter confusion or intentional protest without backing the "No" campaign.
* **Municipal Election Concurrency:**  
  Because the vote occurred alongside contested mayoral races (e.g., Calgary and Edmonton), overall turnout was anchored to municipal issues rather than constitutional mobilization alone.

---

### 5. Strategic Intelligence Takeaways

* **Separate Math from Narrative:** Never mistake high undecision in raw polling for a 50/50 race when decided voter ratios exhibit a 24-point lead.
* **Prioritize Likely-Voter Turnout Screens:** General population panels dilute motivated voting blocs, especially on high-salience constitutional or ballot measures.
* **Track the Protest Option:** In concurrent referendums, evaluate blank and declined ballots as a distinct exit vector for dissatisfied or uncommitted electors.