Election Polling: Minimum Features & Standards

About This Checklist

This checklist outlines the minimum required features for credible, professional election polling. Meeting these standards ensures your poll can be cited by media, included in poll aggregators (FiveThirtyEight, RealClearPolitics), and withstand professional scrutiny. Polls that fail to meet these minimum standards risk being dismissed as unreliable or methodologically flawed.

Introduction to Minimum Standards

Election polling standards exist to ensure:

Methodological rigor - Scientifically sound sampling and analysis
Transparency - Complete disclosure enabling evaluation and replication
Quality control - Procedures to detect and prevent errors
Ethical conduct - Protection of respondent privacy and data integrity

The standards below represent the minimum bar for professional work. Best-in-class pollsters often exceed these requirements significantly.

1. Sampling Requirements

Required Sampling Features

Requirement	Minimum Standard	Acceptable Alternatives
Sample Size (Statewide)	n ≥ 400 likely voters	n ≥ 800 registered voters
Sample Size (National)	n ≥ 1,000 likely voters	n ≥ 1,500 registered voters
Sampling Method	Probability-based (RDD, voter file)	Non-probability with validated calibration weighting
Cell Phone Inclusion	≥ 50% of sample from cell phones	Online sample with mobile-optimized design
Coverage of Population	Frame covers ≥ 95% of target population	Documented coverage adjustments
Multiple Contact Attempts	≥ 5 call attempts (phone surveys)	≥ 3 email invitations + 2 reminders (online)
Field Period	≥ 3 days of interviewing	≥ 2 days if time-sensitive (breaking news)

Critical: Sample sizes below n=300 (statewide) or n=800 (national) are generally considered too small for reliable election polling, regardless of other methodological strengths.

Sample Composition Requirements

Geographic representation: Sample must include respondents from all major regions/areas
Demographic diversity: Adequate representation of age, gender, race/ethnicity groups
No artificial quotas: Avoid hard quotas that terminate interviews mid-stream
Random selection: Within-household selection must be random (not "whoever answers")

2. Likely Voter Screening

Required LV Modeling Features

At minimum, one of the following likely voter screening methods must be employed:

Required question: "How likely are you to vote in the [date] election?"

Minimum response scale: 4-point scale (Absolutely certain, Very likely, Somewhat likely, Not likely)

Likely voter definition: Only include "Absolutely certain" and "Very likely" respondents

Additional requirement: Must verify voter registration status

Minimum questions required:

Voting intention/likelihood (see Option 1)
Past voting behavior: "Did you vote in the [last election]?"
Registration status and location verification
At least one additional turnout indicator (interest, attention, plan to vote)

Scoring: Documented scoring algorithm combining responses

Requirements:

Match respondents to state voter file or commercial voter database
Use validated turnout history from last 2-4 elections
Match rate must be ≥ 80% of sample
Document matching algorithm and success rate
Model turnout probability using past vote history

Best Practice

Report results for BOTH registered voters (RV) and likely voters (LV). This allows readers to see the impact of your turnout model and provides broader context.

3. Weighting Standards

Required Weighting Variables

At minimum, polls must weight to the following dimensions:

Demographic Weighting (Required)

Age (at least 3 categories)
Gender (Male/Female minimum)
Race/Ethnicity (at least White/Non-White)
Education (at least College/Non-College)

Geographic Weighting (Required)

Region/Area (state for national, county/region for state polls)
Population density (Urban/Suburban/Rural) - Recommended

Political Weighting (Highly Recommended)

Party registration (in states with party registration)
Past vote (2020 presidential vote recall or validated vote history)
Vote frequency (always/sometimes/rarely voted)

Education × Race Interaction (Post-2016 Standard)

Following 2016 polling errors, weighting to education within race/ethnicity categories is now considered essential:

White College-Educated vs White Non-College
Non-White College-Educated vs Non-White Non-College

Weight Trimming Required

Weights must be trimmed or capped to prevent extreme weights from destabilizing estimates. Standard practice: trim at 3-5x the median weight. Document trimming procedures and report design effect (DEFF).

Weighting Targets

Acceptable sources for weighting targets:

U.S. Census Bureau: American Community Survey (ACS), Current Population Survey (CPS)
State voter files: Official voter registration databases (for RV samples)
Past election results: Official vote totals for past vote weighting
Exit polls: Validated exit poll demographics (e.g., Edison Research/National Election Pool)

Prohibited: Weighting to current polling averages or partisan preferences. This circular reasoning invalidates results.

4. Transparency & Disclosure Requirements

AAPOR Mandatory Disclosures

The following information must be disclosed for every public release:

1. Sponsorship & Conducting Organization

Who paid for the poll (sponsor)
Who conducted the fieldwork (if different)
Any partisan affiliations or clients
Publicly available contact information

2. Exact Question Wording

Verbatim text of all reported questions
Response options provided
Question order for key items
Any randomization or rotation used

3. Sample Description

Population sampled (LV, RV, adults)
Sample size (n=)
Number of completed interviews
Response rate or completion rate

4. Sampling Method & Frame

Probability vs non-probability
Sampling frame (RDD, voter file, panel, etc.)
Selection procedures
Coverage of target population

5. Interview Mode & Timing

Mode: Phone (live/IVR), online, in-person
Exact field dates (start and end)
Language(s) offered
Average interview length

6. Weighting Procedures

Variables used for weighting
Targets and sources (Census, voter file, etc.)
Weighting method (rake, post-stratify, etc.)
Weight trimming/capping procedures

7. Margin of Error

MOE for full sample (±X%)
Confidence level (typically 95%)
Design effect if applicable (weighted data)
MOE for subgroups if reported

8. Likely Voter Definition

Screening questions used
Criteria for "likely voter" classification
% of sample qualifying as LV
Difference between RV and LV results

Where to Disclose

All required information should be available in at least one of the following:

The published article/press release (abbreviated methodology statement)
A detailed methodology document available on the pollster's website
Direct response to reasonable public inquiry within 3 business days

5. Quality Control Requirements

Mandatory Quality Control Procedures

Interviewer Quality Control (Phone Surveys)

Interviewer training and certification
Call monitoring/supervision (≥10% of interviews)
Validation callbacks (≥10% of completions)
Falsification checks and penalties

Data Quality Control (All Modes)

Speeder detection (completion time ≥ 40% of median)
Straight-line detection in matrix questions
Logical consistency checks
Open-end review for nonsense responses

Online-Specific Quality Control

Bot detection (CAPTCHA, honeypot fields)
Device fingerprinting (prevent duplicates)
IP address monitoring (geo-targeting validation)
Attention checks (instructional manipulation checks)

Data Processing Quality Control

Double data entry or automated validation (CAPI/CATI)
Range and format checks before analysis
Missing data documentation
Audit trail of data cleaning decisions

Data Exclusion Transparency: Any respondents excluded from final analysis (speeders, straight-liners, failed quality checks) must be documented with counts and reasons.

6. Reporting Standards

Required Reporting Elements

Headlines & Summary Statistics

✓ Acceptable:

"Smith 48%, Jones 45% (±3.1%, n=1,000 LV)"
"Within margin of error"
"Statistical dead heat"

✗ Prohibited:

"Smith surges ahead!" (without MOE)
"Clear lead" (when within MOE)
"Landslide victory predicted" (overconfidence)

Required Disclosure in All Public Reports

Sample size and population (e.g., "n=800 likely voters")
Margin of error with confidence level
Field dates
Sponsor/conducting organization
Link to full methodology

Presentation of Results

Report percentages to whole numbers (48%, not 48.3%)
Show both RV and LV results when applicable
Report undecideds and refusals separately
Include trend data from past waves when available
Note any significant changes in methodology from past waves

Subgroup Reporting

Always include subgroup sample sizes (e.g., "Among women, n=450")
Note larger MOE for subgroups
Do not report subgroups with n<100
Mark n=100-200 as "directional only"

Complete Compliance Checklist

Pre-Release Verification Checklist

Before releasing any election poll results, verify that ALL of the following requirements are met:

Sample size meets minimum (≥400 statewide, ≥1000 national)

Sampling method documented (probability or calibrated non-probability)

Cell phones included in sample (if phone survey)

Multiple contact attempts made (≥5 for phone, ≥3 for online)

Field period ≥3 days (or justified if shorter)

Likely voter screening implemented

Likely voter model documented and validated

Registered voter results also available

Response/completion rate calculated

Weighted by age (minimum 3 categories)

Weighted by gender

Weighted by race/ethnicity

Weighted by education

Weighted by geography (region/area)

Education × race interaction weighting (post-2016 standard)

Weights trimmed/capped at reasonable threshold

Design effect (DEFF) calculated and acceptable (<2.0 preferred)

Sponsor clearly identified

Conducting organization identified

Exact question wording available

Sample description complete (size, population)

Sampling method disclosed

Interview mode and field dates disclosed

Weighting procedures documented

Margin of error calculated and reported

Speeder detection implemented

Straight-line detection for matrix questions

Data quality exclusions documented

Fraud/bot detection (online) or validation (phone)

Missing data handled appropriately

Data processing audit trail maintained

MOE included in headline/summary

Undecideds and refusals reported separately

Subgroup sample sizes disclosed

Avoid overclaiming statistical significance

Full methodology document publicly available

Checklist Complete?

If you can check ALL 36 items above, your poll meets professional standards for public release. If any item is missing, address it before publication or clearly note the limitation in your disclosure.

PollZapper: Built-In Compliance

PollZapper automatically ensures compliance with all minimum standards:

Automated AAPOR disclosure generation
Pre-configured likely voter models
Standard weighting to Census targets
Automatic MOE calculation

Built-in quality control flagging
Weight trimming and DEFF reporting
Complete audit trails
Publication-ready methodology statements

Join Waitlist for Early Access

Related Resources

Election Polling Guide

Comprehensive guide to election polling standards and methodology.

View Guide

Sampling Methods

Learn about probability sampling, weighting, and sample size calculation.

View Sampling Guide

Question Types

Best practices for designing effective survey questions.

View Question Types