Questionnaire and Sample Design
The first phase of the survey research process is the creation of a valid and reliable survey instrument. Our project management team is well versed on the best practices of questionnaire design, and we pay careful attention to all the elements of the survey: topics and content, length and format, and question wording and order. By collaborating closely with our clients to clearly identify and define study objectives, we create surveys with relevant questions and understandable response categories. Before a survey goes into the field, we conduct numerous pre-tests of the survey instrument, to eliminate potential problems with question wording and sequence, awkward terminology, or confusing response categories. This testing allows for necessary changes and ensures that we collect high-quality data. Whether you are starting from scratch, or have an existing survey that needs some fine-tuning, we have the knowledge and expertise to facilitate this phase of your research project.
The second phase of a survey involves sample design. Once we know your project's goals and budget, we can help you choose the best population-or sample-to survey. By choosing a sample that appropriately represents the population you want to study, we can enhance the reliability and validity of the survey data. Currently, we use the following sampling and screening methods, choosing the ones that will best meet the survey's goals.
Random Digit Dialing
For this sampling technique, a computer program generates a list of telephone numbers from known area codes and prefixes. Using the existing exchanges, the program randomly creates the final four digits to be dialed. This is the most efficient means of sampling households, and provides for an equal and known probability of selection for all residential telephone numbers in the area selected for study. Random digit dialing has been used by survey researchers for many years, and has proven an extremely reliable sampling methodology, producing results that accurately reflect true population values.
For this method, a random sample of households in the area to be surveyed is created using the U.S. Postal Service's Delivery Sequence File (DSF). Because the DSF includes more than 95% of households, this provides a way of reaching residents whether they have a landline telephone or not. Using the DSF, known household addresses are matched to telephone numbers where possible, and households with a matched telephone number are mailed a postcard informing them that an interviewer will be calling. Households without a matching telephone number are sent a packet of materials, sometimes including a small monetary incentive, asking them to contact the Center for Opinion Research to participate in the survey.
If the population to be surveyed is very specific, then a client will provide us a list of names with addresses, phone numbers, and/or email addresses. Using this list, we will either develop a random sample or attempt to reach all the people on the list, depending on the goals of the survey. List samples are useful for reaching defined groups such as members of an organization, registered voters, recent patients of a hospital, parents and guardians of public school students, or recent graduates of a university. List samples are also used for recruiting participants for other kinds of studies, such as focus groups or depth interviews.
For telephone surveys, further screening techniques, or second stage sampling methodologies, are used to eliminate biases that might arise from simply interviewing the person who answers the telephone. These include the Last and Next Birthday Methods, where an interviewer asks to speak with the adult living in the household who most recently celebrated, or will next be celebrating, a birthday, and specific screeners, such as asking to speak with the person in the household who makes the healthcare decisions or who is the parent or guardian of a school-aged child living in the home. Using these secondary screening techniques greatly enhances the validity of survey data.