THE INDEPENDENT AND NON-PARTISAN SURVEY OF PUBLIC OPINION ESTABLISHED IN 1947 AS THE CALIFORNIA POLL BY MERVIN FIELD   601 California Street San Francisco, CA 94108 (415) 392-5763   FAX: (415) 434-2541 info@field.com

The Survey Methods Used by The Field Poll Over the past thirty years, The Field Poll has typically employed a random digit dial sampling methodology when conducting surveys of either California adults or the state's registered voter population. This sampling method has been, and continues to be, the industry standard when conducting surveys of adult populations. However, beginning in 2006, when conducting surveys of the state's registered voter population, The Field Poll has employed registration-based lists of the state's voting population as its sample frame. When conducting surveys of registered voters from either source, the goal is the same; namely, to develop a representative sample of the overall registered voter population in California. Description of Sampling Procedures The following is a summary of the procedures used by The Field Poll when using a random digit-dial (RDD) sampling approach and those employed using a registration-based sampling (RBS) methodology. A. Sampling Using a Random Digit Dialing Methodology When Field Poll samples of the California telephone household population are developed using a random digit dialing (RDD) technique, the sample lists are typically generated by Survey Sampling, Inc., a leading supplier of RDD samples to the public opinion research industry. The RDD selection procedure starts with the identification of all telephone exchanges that serve that jurisdiction. Non-working and business blocks are identified in each exchange and eliminated. After working exchanges and blocks of numbers have been weighted proportionally by the number of listed phone lines, numbers are systematically selected to yield a probability sample of base numbers. Finally, random digits are added to the base number to create a random ten-digit telephone number. This is done in the following manner: To equalize the probability of telephone household selection from anywhere in the area sampled, samples are first systematically stratified to all counties in proportion to each county's share of telephone households statewide. To obtain reasonable estimates of telephones by county, a special data base was developed, beginning with Census data for residential telephone incidence. These counts, updated yearly with data on new telephone installations provided at the state level, are then applied to current projections of households by county. After a geographic area has been defined as a combination of counties, the sum of the estimated telephone households is calculated and divided by the desired sample size to produce a sampling interval. A random number is drawn between 0 and the interval (125) to establish a starting point. Assuming the starting point is 86, then the 86th, 211th, 336th, 461st, etc., records would be selected for the sample, each time stepping through the data base by a factor of 125. This is a systematic random sample - as the sample is selected in a systematic "nth" fashion from a random starting point. Any county whose population of estimated telephone households equals or exceeds the sampling interval is automatically included in the sample, while smaller counties are included with a probability proportionate to their size. For each county included in the sample, one or more unique telephone numbers are selected by systematic sampling from among all working blocks of numbers in all telephone exchanges assigned to the county. A working block is defined as 100 contiguous numbers containing three or more residential telephone listings. In this example, for the exchange 226, the entire block comprises the numbers 7500-7599. Exchanges are assigned to a single county on the basis of where listed residents live. Nationally, about 80 percent of all exchanges appear to fall totally within county boundaries. For those overlapping county lines, the exchanges are assigned to the county with the highest number of listed residents. Once the sample has been allocated, a second sampling interval is calculated for each county by dividing the number of listed telephone households for the county by the portion of the sample allocated to that county. Each exchange and working block within an exchange is weighted by its share of listed telephone households. If the total number of listed telephone households in the data base for this county is 45,000 and that county is allocated 500 numbers, the resulting interval produced would be 90. Next, from a random start between 1 and 90, those exchanges and working blocks falling within the interval are sampled on a systematic basis. Two or more digits randomly chosen from the range 00-99 are then added to each of the blocks selected. The result is a complete number made up of the exchange, the block, and the two random digits. B. Sampling from Registration-based Lists When Field Poll samples of registered voters are implemented using a registration-based sampling (RBS) methodology, lists of registered voters statewide are purchased from Voter Contact Services, a leading supplier of voter lists to the survey research industry. The list is updated regularly and includes the names of virtually all registered voters statewide, along with a wealth of other information about the voter, including a voter's address, city and county of residence, gender, date of birth (age), party registration, whether or not the voter is a permanent absentee voter, and extent to which the voter participated in past elections. The list currently provides a telephone number for over 90% of the voters listed. These telephone numbers come from a variety of sources, including telephone numbers included on the voter's registration form, as well as by cross referencing voter names and addresses against recent telephone directories and other telephone matching services. While the telephone numbers included are primarily landline telephone numbers, these phone listings include cell phones, whenever a cell phone number is provided by the voter when registering to vote or in other settings accessible to telephone matching services. Thus, cell phones are not systematically excluded from the sample frame. Special procedures are employed when contacting voters on their cell phones to include them into each survey. Another difference when sampling from an RBS list is that the RBS sample frame is drawn from a list of individuals (i.e., voters), whereas the sampling frame for RDD surveys is a household (through the identification of household landline phones identified when implementing the RDD sample). Because of this, the RBS sampling approach eliminates the need to implement any respondent selection procedures once a contact is made, because the full name of the voter to be sampled is known in advance. In addition, RBS sampling does not rely on respondent testimony as to whether the adult being sampled is a registered voter, and if so, what party the voter is affiliated with, since all persons contacted are known to be registered and their actual party registration is identified. When drawing samples from the RBS database, The Field Poll employs a random selection procedure giving all voters with telephone numbers an equal chance of being selected. All interview attempts are made only with the specific voter selected. Since the voter's gender, county of residence, age and party registration are known in advance additional samples of voters meeting specified sampling requirements can easily be augmented or the statewide sample stratified along any of these dimensions. Because the name and address of the voter is known, RBS samples can also be coupled with either an advance letter or a follow-up letter mailed directly to the targeted voter, informing him or her that a telephone interviewer will be or has tried to include them in a statewide Field Poll survey. Letters can be sent to all or parts of the voter sample (e.g., follow-up letters can be sent to voters who initially refused to participate in the survey, or to those in which the initial attempt yielded a cell phone message.) These letters, printed and mailed on Field Poll letterhead, serve to provide voters with information about the survey's auspices, help in distinguishing the survey from unsolicited telemarketing or political calls, and provide a toll-free telephone number for Field Research Corporation to allow voters to participate in the survey on a day and time of their own choosing. Conversion of the Survey Questionnaire Onto Field's CATI System After a penultimate version of the questionnaire has been developed for each survey, it is translated into Spanish and both the English and Spanish language questionnaires are programmed onto Field Research Corporation's computer-assisted telephone interviewing (CATI) system in preparation for testing. CATI controls the telephone scripts read to individual respondents by displaying the appropriate questionnaire items and their valid response code alternatives in their proper sequence on computer screens at each interviewer's booth. The interviewer then reads each question aloud to the respondent from the screen and enters the pre-coded answer category through the keyboard directly to a computer disk. All answers are automatically stored in computer memory. Online interviewing using CATI allows for greater consistency in interviewing by controlling skip patterns, branches, randomization of items in a battery, "refer backs," and other control features during the call. CATI also affords greater opportunities for internal control, since the development and programming of the questionnaires remain under the direct control and supervision of the project director. This ultimately helps to insure that all interviewing procedures and scripts are set-up and implemented in an identical fashion for each respondent. In addition to sequencing and personalizing questions, the CATI program performs various quality control functions, including on-line editing. The program will reject ineligible codes entered by an interviewer to all pre-coded questions. The CATI script is prepared for a telephone pre-test among a small sample of adults to assess general ease of administration, refine item wording and provide an initial assessment of average length of administration. Following this, changes are made to the questionnaire, as appropriate, and these are incorporated into the CATI script in preparation for a more formal pilot test Data Collection Procedures Telephone interviewing for each Field Poll survey is typically conducted internally drawing from Field Research Corporation's large corps of trained interviewers, with full-time staff professionals on hand to supervise, monitor and evaluate the performance of each interviewer. Field Research Corporation's interviewing facilities consist of sound-protected booths where interviewers are stationed to do the calling using state-of-the-art computer-assisted telephone interviewing. In order to bring hard-to-reach respondents into its surveys, up to eight attempts (an initial call plus seven callbacks) are made to each telephone number selected for inclusion into the sample. Callbacks are made at different times and on different days to increase the probability of finding adults available for the interview. Where possible, appointments are made at specified dates and times to maximize convenience and cooperation rates. Interviewer Training Procedures The role of the interviewer is critical in obtaining accurate and reliable survey data. Consequently, interviewers working on each Field Poll study are carefully trained in all the nuances of questionnaire administration and monitored throughout the interviewing period to assure uniform practices. The following measures are employed to assure high quality and uniform telephone interviewing practices: All interviewers working are required to complete an interviewer training course, which will provide both general and specific interviewing instructions, refresher reviews and on-line monitoring of telephone interviewing. During the training course, interviewers are provided with an interviewer training manual. The training course and manual includes an introduction to survey research, a description of interviewer roles and responsibilities, general interviewing techniques and record keeping, refusal conversion techniques, and confidentiality procedures. In addition, procedures are reviewed for the proper management of non-English speaking households. At the conclusion of their training, interviewers conduct mock interviews, and their performance is evaluated by Field Research Corporation's professional interviewing supervisors. Before the start of data collection, all interviewers working on the study are required to attend a briefing session where the calling and interviewing procedures are described in detail by the Study Director. This session provides both interviewers and supervisors with an overview of the study and includes a question-by-question review of all items in the survey. The session also discusses recommended best-practice approaches for dealing with different interviewing situations, documenting the results of contact attempts, scheduling of callbacks and confidentiality requirements. Debriefings and retraining sessions are held as necessary to be sure that all interviewers are following consistent procedures. The performance of each member of the interviewing team is closely monitored and evaluated throughout data collection. In addition, from time to time interviewers are asked to meet together as a group to discuss their interviewing experiences on the project. If questions arise and clarifications need to be made about specific survey questions, written responses are prepared and distributed to the interviewing staff to ensure uniform, standardized interviewing practices. Throughout the interviewing period "data correction sheets" are available to all interviewers to note respondent changes to answers after the initial recording of their responses during the interview. Data Processing Procedures All processing and survey tabulations are generated from Field Research Corporation's own fully equipped, in-house data processing facility. This allows for close supervision and control of all processing functions by the Study Director and other project team members. Basic tabulations are handled with SPSS MR Ltd.'s Quantum software package, a highly efficient and sophisticated cross-tabulation package for survey research data. The following is a description of the procedures that Field Research Corporation employs to complete the data processing tasks for each Field Poll study. Data File Preparation All information derived from the survey is systematically formatted in preparation for data cleaning and processing. Post-Interview Coding Tasks Survey questions which permit verbatim replies to open-ended questions are coded and key-entered into each respondent's data file by Field Research Corporation's coding staff after the completion of data collection. Approximately twenty percent of the replies to each open-ended question are sampled by full-time coding staff. Using the sampled responses, the Study Director and the Coding Supervisor establish tentative code categories to permit detailed coding and quantification of all qualitative responses. Each coder's work is checked to maintain accuracy and consistency in the coding effort. Edited and coded questionnaires are then keyed into Field Research Corporation's in-house computer. Data "Cleaning" and File Checking Because CATI itself provides for the direct data entry of responses by the interviewer and does not permit ineligible or invalid data entries, the data file resulting from all CATI interviewing is itself virtually error-free. However, because interviewers manually fill out data correction sheets when a respondent changes his or her response after it has been entered, the survey data require additional data "cleaning." All data correction sheets are reviewed and interview information corrected, as necessary, by Field Research Corporation's professional coding staff. Following this, an additional series of checks are performed by means of a specially designed cleaning program that will scrutinize each questionnaire for internally inconsistent information. Weighting To generalize survey data to the overall population of California adults or voters, statistical weights are developed to adjust for possible deviations in the probability of selection of households and individuals, and to account for minor variations in the representation of individual demographic subgroups of the population that result from the sampling process. Field Research Corporation's particular weighting software is quite sophisticated and internally calculates individual weights from specified "target values" established by the Study Director. For example, in developing weights for the adult sample by age and gender, the researcher simply specifies the proportion of the total target population who men age 18-24, the proportion who women age 18-24, and so on as the target weight values desired in the final weighted sample. This greatly simplifies both the construction and proofing of the weights, since the correct target values can be displayed and checked in each weighted table at the end of the weighting process. Processing and Tabulation At the conclusion of this process, a clean data file is prepared and from it detailed statistical tabulations are produced. These tabulations display the results of each survey question overall statewide and across a set of regional and demographic subgroups of the state's population, and form the basis from which Field data analysts prepare and present survey results in its reports and press releases. Reporting Reports of The Field Poll, typically published 30-40 times per year, cover a wide range of political, social and economic topics. Continuing measures are made of voter support for leading political figures vying for major state and federal elected offices, job ratings of important political figures, and reactions to significant political events. Voter awareness, understanding and predispositions for major campaign issues and controversial ballot propositions are also tracked over time. Each Field Poll release consists of two to twelve pages of text and statistical data presented in press release format, plus a background fact sheet reporting the details of the survey, its sample size, exact question wording and other technical details from that particular poll. Many of California's leading media properties and news organizations publish and broadcast the results of each Field Poll report as feature or breaking news stories in their respective publications and broadcasts. All new and previous reports since 1995 are available to the public through our online archives. Academic institutions, including the University of California and California State University systems, have access to source data files and codebooks in SPSS format from each survey, typically within ninety days of the completion of the study. These data files provide each campus with the ability to access ongoing contemporary public opinion data suitable for teaching and research. The Field Poll is owned and operated by Field Research Corporation, with headquarters in San Francisco, California.

Back to Home Page