Did precincts with high Democratic registration support the school bond? How closely did the local candidate's performance track with the top-of-ticket race? Which precincts split their tickets? The Precinct Analysis tool answers these questions by comparing any two electoral variables across all precincts in a jurisdiction, calculating correlation statistics that quantify the relationship.
What Correlation Analysis Reveals
Correlation analysis tests hypotheses about voter coalitions:
- Partisan alignment: Did a ballot measure or candidate's support track with party registration?
- Ticket-splitting: Where did voters support one party's Presidential candidate but another party's local candidate?
- Demographic patterns: Did precincts with high turnout from specific ethnic groups correlate with support for particular candidates?
- Issue coalitions: Did two different ballot measures draw from the same voter coalition or different ones?
The tool doesn't prove causation—it shows whether patterns exist and how strong they are.
Finding Precinct Analysis
On any jurisdiction page:
- Click the Elections tab
- Select Precinct Analysis
The tool is only available for jurisdictions with precinct-level data.
Setting Up a Comparison
The interface has two selection panels:
Selection A and Selection B each allow you to choose:
- A category (Party Registration, Party Turnout, Ethnic Registration, Ethnic Turnout, or a specific contest)
- A specific option within that category (e.g., "Democratic" for Party Registration, or "Yes" for a ballot measure)
Use the controls at the top to set:
- Year: The election year to analyze
- Election type: General or Primary
- Min. Ballots: Exclude small precincts below a threshold (reduces statistical noise)
Reading the Results
After selecting both variables, the tool displays:
Correlation Coefficient: A number from -1 to +1
| Range | Interpretation |
|---|---|
| 0.8 to 1.0 | Strong positive—precincts high on A tend to be high on B |
| 0.5 to 0.8 | Moderate positive |
| 0.2 to 0.5 | Weak positive |
| -0.2 to 0.2 | No meaningful correlation |
| -0.2 to -0.5 | Weak negative |
| -0.5 to -0.8 | Moderate negative |
| -0.8 to -1.0 | Strong negative—precincts high on A tend to be low on B |
Vote Share: Overall percentage for each selection across the jurisdiction.
Precincts Higher: How many precincts had Selection A higher than Selection B (and vice versa). A selection can win more precincts but have lower overall vote share if votes are concentrated differently.
Average Difference: The mean percentage point gap between selections across all precincts.
The Precinct Table
Below the summary statistics, a table lists every precinct with:
- Selection A percentage
- Selection B percentage
- Difference (positive = A higher, negative = B higher)
- Total ballots cast
Click column headers to sort. Click "Details" on any row to see that precinct's demographic composition and a mini-map of its location.
Toggle between "Percentages" and "Raw Numbers" views depending on your analysis needs.
Example Analyses
Was the ballot measure partisan?
Compare: Party Registration → Democratic vs. Proposition X → Yes
- Strong positive correlation (0.7+): The measure tracked closely with partisan registration—Democratic precincts voted Yes, Republican precincts voted No
- Weak or no correlation (under 0.3): The measure cut across party lines—both parties' voters split on it
Where did the candidate outperform their party?
Compare: Presidential Race → Democratic Candidate vs. Local Race → Democratic Candidate
Sort the table by "Difference" to find precincts where the local candidate ran ahead of or behind the Presidential candidate. Precincts with large positive differences show the local candidate's personal vote; large negative differences show weakness relative to party baseline.
Did ethnic composition correlate with candidate support?
Compare: Ethnic Turnout → Hispanic/Latino vs. Assembly Race → Candidate A
A strong correlation suggests Candidate A's support was concentrated in precincts with high Latino turnout. A weak correlation suggests other factors mattered more, or that Latino voters were split.
Did two ballot measures draw from the same coalition?
Compare: Proposition X → Yes vs. Proposition Y → Yes
High correlation means voters who supported one tended to support the other. Low correlation means the measures appealed to different coalitions, even if both passed or failed.
Important Limitations
Correlation is not causation. A strong correlation between Democratic registration and Yes votes doesn't prove Democrats voted Yes. Other factors correlated with Democratic registration might be the actual driver.
This is ecological data. You're comparing precinct-level aggregates, not individual voters. "Precincts with more Democrats voted Yes" is not the same as "Democrats voted Yes." This is a well-known limitation called the ecological fallacy.
Small jurisdictions produce unreliable correlations. With only 15 precincts, apparent correlations may be statistical noise. Be skeptical of strong results in small jurisdictions.
Ethnicity is modeled. Ethnic categories are derived from surname matching, which has accuracy limitations. Don't over-interpret small differences between ethnic groups.
Comparisons are within a single election. Precinct boundaries can change between elections. Cross-year comparisons require caution.
Practical Applications
Testing campaign hypotheses: "We won because of strong Latino support." Run the correlation to see if the data supports this—or reveals a different story.
Identifying persuadable precincts: Find precincts where partisan registration didn't predict the outcome. These show where voters crossed party lines.
Understanding issue coalitions: Before launching a ballot measure campaign, check how similar past measures correlated with party registration. A non-partisan correlation pattern suggests different messaging and coalition-building strategies than a partisan one.
Campaign debrief: Quantify how closely a candidate's performance tracked with expected baselines. Did they run a partisan race or build a cross-partisan coalition?
Common Mistakes
Confusing correlation strength with percentage point difference: A correlation of 0.9 means the pattern is highly consistent, even if the average difference is only 2 percentage points. Correlation measures relationship consistency, not magnitude.
Ignoring the Min. Ballots filter: Tiny precincts with 30 voters can have wild percentage swings that distort correlations. Filter them out.
Over-interpreting single correlations: One comparison can't tell the full story. If Democratic registration correlates with Yes votes, check whether the correlation is unique or simply reflects that Democratic precincts also have other characteristics that drove the vote.
Expecting perfect correlations: Even Democratic registration vs. Republican registration rarely exceeds -0.95 correlation (because "Other/NPP" voters exist). Correlations above 0.85 are very strong.
Comparing across different election types: Primary election turnout composition differs dramatically from General elections. Compare like to like.