Morning Scrapple Methodology

Overview

Morning Scrapple is a recurring survey of eligible Pennsylvania voters. The survey utilizes a non-probability, multi-source recruitment strategy that includes:

• • • Text messages sent to matched phone numbers from the Pennsylvania voter file
    • Email invitations to opt-in participants previously recruited from the voter file and digital advertising
    • Digital advertising, including social media placements that allow respondents to see and take the survey

Non-probability samples are not described using a traditional margin of sampling error.

Sampling & Recruitment

Respondents are recruited by:

• • • Text messages to voter-file–matched numbers
    • Email invitations to a recontact list composed of prior opt-ins
    • Digital advertising served statewide (e.g., Meta), allowing voluntary participation

Respondents are offered modest incentives that vary by recruitment channel. Participants recruited via paid digital advertising receive a $5 instant digital gift card of their choice upon completion, while participants recruited by text or email are entered into a randomized drawing for a $100 digital gift card. Incentive structures encourage diverse participation without conditioning responses and are disclosed in accordance with AAPOR transparency standards.

Recruitment sources are disclosed in the topline report along with their percentage contribution to the final sample.

Field Period

• • • Surveys are fielded online over a defined multi-day window.
    • Completion rates, field dates, and sample disposition (AAPOR outcome codes) are published with each survey release.

Weighting

To improve representativeness, data are weighted using calibration weighting or propensity-adjusted ranking to match population benchmarks derived from high-quality public and administrative sources, including the Pennsylvania voter file and U.S. Census–based estimates. The minimum adjustment set includes:

• • • Age
    • Gender
    • Geography (state, county or region)
    • Party registration

Weights are examined for outliers and trimmed at the top and bottom percentiles to reduce variance inflation. The final report includes:

• • • Weight distribution summary
    • Design effect (DEFF)
    • Effective sample size (n_eff), which is used for uncertainty estimation

The recruitment source is monitored for composition effects but is not included in the weighting scheme, as AAPOR guidance cautions against adjusting for variables without reliable population benchmarks or demonstrated explanatory power beyond core demographic and political controls.

Uncertainty Estimates

Per AAPOR guidance, Morning Scrapple reports a 95% Credibility Interval.

A model-based estimate of uncertainty derived from the weighted data and the effective sample size.

• • • Represented as:
      “95% credibility interval: ±X.X percentage points.”

This interval does not imply random sampling and reflects both sampling variability and uncertainty produced by weighting adjustments.

Questionnaire & Transparency

• • • Full questionnaire wording and response options are published alongside the topline results.
    • Sponsorship and fieldwork details are disclosed in accordance with AAPOR’s Transparency Initiative.
    • Fraud detection protocols (device duplication, speed checks, IP verification) are described, including counts of removed or flagged responses.

Data Quality

The results represent only the opinions of individuals who chose to participate.
Quality controls include:

• • • Speeding checks
    • Duplicate response screening
    • Device- and browser-level validation
    • Consistency checks for straight-lining or patterned responding

Rejected response counts are reported.

Disclosure Statement

• • • Respondents are recruited using a non-probability, opt-in sample consisting of SMS invitations, email recontacts from a voter-file–derived list, and statewide digital advertising.
    • The data were weighted using calibration/propensity-adjusted methods to align the sample with Pennsylvania adults on age, gender, and party registration.
    • Because the sample is not a probability sample, a traditional margin of sampling error does not apply.
    • A model-based 95% credibility interval is provided to illustrate approximate uncertainty after weighting.
    • Results reflect only the views of those who elected to participate in the survey.

Field dates, sample composition, response metrics, and survey questions are publicly available.

Outcome Codes