Methodology

The Marist Poll uses a rigorous, scientific-based,

dual-frame approach in our telephone survey methodology.

General Information

The Marist Poll uses a rigorous, scientific-based, dual-frame approach in our telephone survey methodology.

Samples are designed to represent the adult population, aged 18 and older. Surveys are conducted at the national, statewide and citywide level. In-house, using Dyanta, a geographically stratified, random-digit-dialing (RDD) sample of telephone households in the targeted calling area is selected. Telephone numbers are chosen based upon a list of known telephone exchanges from throughout the target area. The exchanges are selected to ensure that each region (strata) is represented in proportion to its population. All sample telephone numbers are computer generated and managed by our computer-assisted telephone interviewing (CATI) system.

Currently, more than half of our national sample is obtained from a mobile telephone frame. Proportions may be adjusted in state and city projects based on cell-phone-only populations from the most current National Health Interview Survey (NHIS) Wireless Substitution Estimates. Mobile phones are treated as individual devices. After validation of age, personal ownership, and non-business-use of the mobile phone, interviews are typically conducted with the person answering the phone.

To increase coverage, the Marist Poll includes the dialing of landline phone numbers. Within each landline household, a single respondent is selected through a random selection process to increase the representativeness of traditionally under-covered survey populations.

The Marist Poll collects data using our state-of-the-art survey center located in the Hancock Center at Marist College’s Poughkeepsie, New York campus. Data collection is typically conducted over a three-to-five-day period depending on sample size, length of questionnaire, and incidence of target population. Due to current regulations, phone sample is manually dialed and called by live interviewers. The result of every call attempt (e.g., completed interview, answering machine, refusal, disconnect) is documented so that re-contacts may be attempted according to strict “callback rules.”

VOXCO, the Marist Poll’s CATI system, allows for exceptional control of sample management, questionnaire administration, seamless handling of skip patterns, data quality checks, and data processing.

The Marist Poll employs more than 400 undergraduate students each semester who,after extensive in-house training, serve as researchers and telephone interviewers. Interviewers are hired after a pre-assessment process and then trained in person either online or on-site. Training includes discussion of the protection of human subjects and CATI procedures. Prior to the start of every study, interviewers are trained and briefed on issues specific to a particular project. Each question is reviewed to promote interviewer consistency. Throughout data collection, strict data quality control procedures are maintained. Field supervisors continually monitor the interviewers and incoming data using both aural and visual validation methods. Each time an interview is monitored, a supervisor/coach provides immediate feedback to encourage growth and success. Online evaluation forms are completed by supervisors/coaches to assess interviewers on specific aspects of the data collection process. In addition, all calls are recorded and stored on a high-security server.

Post data collection, raw data from both landline and cell phone frames are extracted, cleaned, and combined for balancing to population parameters and analysis using a suite of data management and statistical packages (e.g., SPSS Statistics, MarketSight). The Marist Poll post-stratifies and balances its data to reflect Census results for age, gender, income, race, and region.

Data is typically reported for adults as a whole (topline), as well as by subsets of interest (e.g., age, race, gender, political party affiliation, income, education). For studies including hypothetical candidate preference questions, results may be reported for registered and/or likely voters. The Marist Poll uses a probability turnout model to identify likely voters. This model determines the likelihood respondents will vote in the current election based upon their chance of vote, interest in the election, and past election participation. It should be noted that the Marist Poll does not weight its data by party identification.

Error margins are reported for topline data and key subsets. The error margin is adjusted for sample weights and increases for cross-tabulations.

As a member of AAPOR’s Transparency Initiative, the Marist Poll adheres to and encourages industry standards of methodological transparency. Please feel free to contact us if you have additional questions about our telephone survey process.