How we do this

This page describes the standing methodology that applies to every Texas Policy Polling release unless a specific wave deviates and documents the deviation on the release itself.

Sample frame

Respondents are drawn from the Texas voter file. We use the L2 Political voter file (refreshed quarterly) as the universe. For statewide waves the sample frame is the full active registered electorate; for district-level waves it is the active registered electorate within the district.

Voter-file matching

Each respondent is matched to their voter file record at the time of survey invitation. This match is used to (a) verify residency in the intended sample frame, (b) anchor demographic and vote-history features for analysis and weighting, and (c) enable post-fielding district and county cuts.

We disclose this matching practice in our privacy policy and at the start of every survey instrument. Respondents who do not wish to participate under those terms can decline; we honor opt-out requests against the underlying record so the voter is not invited to future waves either.

Recruitment

Invitations are issued by email, sent from texaspolicypolling.orgvia a dedicated sending infrastructure with industry-standard authentication (SPF, DKIM, DMARC) and RFC 8058 List-Unsubscribe headers. Each invitation includes a personalized URL keyed to the respondent's voter record. Reminder sends are limited to two per wave.

Instrument design

Instruments are designed to elicit honest signal on the topic of inquiry without leading respondents to a predetermined answer. Question wording follows the principle that the same instrument could be fielded by an opposing campaign without producing materially different results. Skip logic is used where appropriate (for example, vote-recall questions are conditional on stated vote participation).

Weighting

Raked weights are applied against four targets drawn from the voter file: age band, county, party primary participation in the most recent cycle, and gender. Weights are documented per release.

Reporting

Every release publishes topline results, key crosstabs, sample size (N), fielding window, margin of error, weighting targets, and any known biases or limitations. Where a wave is sponsored by a paying client, the sponsor is disclosed on the release.

Limitations

Email-recruited panel research carries known biases: under-sampling of voters without deliverable email, over-sampling of high-engagement mailbox providers, and self-selection bias relative to random-digit dial. We disclose mailbox-provider mix on each release and weight against voter-file demographics rather than the raw respondent pool to mitigate selection effects. Some bias is irreducible; we are transparent about the residual.

Replication

Survey instruments, weighting code, and crosstab definitions are available on request for any release older than 90 days. Email research@texaspolicypolling.org.