Telephone Surveys
Telephone surveys are an efficient way to collect and compile data. Technology aimed at increasing the efficiency and cost-effectiveness of telephone surveys has advanced a great deal over the past several years. At the ESRL, telephone surveys are conducted using computer-assisted telephone interviewing, which greatly enhances the efficiency and accuracy of collecting data via telephone. To learn more about how telephone surveys are conducted, click on the commonly asked questions below.
- What are the advantages of telephone surveys over other methods of collecting data?
- What kinds of questions are asked in a telephone survey?
- How do you decide who is interviewed?
- Is the information gathered in a telephone survey accurate?
- How do you collect the data?
What are the advantages of telephone surveys over other methods of collecting data?
Compared to other methods of collecting survey data, telephone surveys offer a high degree of efficiency. When a respondent is on the phone, we record their answers to our questions right then as opposed to waiting for them to complete and return a paper questionnaire. We are also not restricted by geography or physical barriers with a telephone survey, which further increases efficiency.
Another advantage of telephone surveys is the ability to incorporate branching. For example, if the respondent says they are not married, we obviously do not want to ask them the age of their spouse. Branching, even when complex, is seamless and invisible to the respondent in a telephone survey but not in other methods.
A final advantage of telephone surveys over other methods of data collection is that telephone surveys typically enjoy a higher response rate than other methods, particularly self-administered questionnaires such as mail or web surveys.
What kinds of questions are asked in a telephone survey?
While the topics covered in a telephone survey vary widely, the way in which the questions are asked do not. There are two basic question formats, called closed-ended and open-ended.
A closed-ended question has predetermined answer categories that each respondent's answer must fit into. Sometimes one of the categories is "other" in case an answer does not fit any other category. An example of a closed-ended question is if we asked the respondent's gender. Obviously the answer categories will be "male" and "female."
An open-ended question does not have predetermined answer categories, but instead allows the respondent to use their own words to answer the question. An example would be if we asked a respondent to describe the main duties of their job. When an open-ended question is asked, the response is typed verbatim by the interviewer and recorded with all the other data. Answers to open-ended questions can later be coded numerically for quantitative analysis if desired.
How do you decide who is interviewed?
The goals of the project dictate who will be interviewed in any survey. For results that can be generalized to a larger population, a random sample is called for. With a telephone survey, this means using random digit dialing (RDD) sampling methodology. A set of randomly-generated phone numbers from the area in which your population of interest resides (which could be anywhere from a few city blocks to the entire nation) serves as the sample. Many of these phone numbers, of course, will lead to businesses or other non-households and many will be disconnected or otherwise not in service. However, an RDD sample yields a representative sample as opposed to a sample drawn from a telephone book, which excludes households with unlisted numbers (a significant proportion of households) and is always at least somewhat out of date.
When a household is reached in a survey using an RDD sample, we typically select an eligible member of that household at random to participate in the survey. This helps ensure that we are indeed surveying a representative sample from the population. If, on the other hand, the survey is only relevant to certain kinds of people in the population we can also screen out those who do not qualify at the respondent selection stage.
While an RDD sample is usually best when the results must be generalized to a larger population, other kinds of samples are also possible. For example, a sample can be generated from a membership list or directory for projects involving a targeted or limited population.
Is the information gathered in a telephone survey accurate?
While the use of computer-assisted telephone interviewing helps ensure a high degree of accuracy in the recording of responses, the ability to accurately generalize survey results to a larger population depends on the response rate of a survey. In any telephone survey, some phone numbers are called and repeatedly result in no answer or an answering device and some people will decline the invitation to participate. The problem with this non-response is that these people, if they could have been interviewed, might have provided answers that differ from those who were interviewed. They also might differ from respondents on important demographic characteristics. Therefore, we attempt to minimize non-response in telephone surveys.
One method of reducing non-response, and thus increasing the accuracy of the data, is to make as many attempts as possible at reaching potential respondents. These call attempts can not only be great in number, but are usually spread out throughout the course of a study so we avoid attempting the same "dead end" at the same time of day or on the same day of the week over and over again.
In order to minimize the proportion of the sample who decline the invitation to participate, we train interviewers to listen for clues as to why the person is not participating and to tactfully encourage a change of heart. We also choose our words carefully in introducing the survey to potential respondents and allow those who initially decline the invitation to participate subsequent opportunities to change their mind.
We are also capable of conducting interviews in Spanish for respondents who are more comfortable speaking in that language. For such instances, a Spanish version of the survey is available at any time to those on our interviewing staff who are fluent in Spanish.
How do you collect the data?
Data for telephone surveys are collected from our centralized phone bank. Our position on the university campus allows us not only to operate as a centralized unit, but also gives us access to faculty with expertise in a wide range of disciplines and research interests, as well as a readily available staff of student and non-student interviewers.
The ESRL uses WinCATI telephone interviewing software from Sawtooth Technologies. WinCATI provides a great deal of flexibility in how the survey is conducted, from allowing complex branching to changing the appearance and even language of the survey. It also allows control over the management of the sample, including when calls are placed and how certain call outcomes are handled.