CHAPTER 5

Modes of Survey Delivery

All research starts with questions. A survey is one way to get the information we need to answer these questions. In order to carry out a survey, we have to deliver the survey to our potential respondents. In this chapter, we’re going to discuss four basic modes of survey delivery. We’ll discuss approaches that combine features of two or more of these modes later in this chapter when we discuss mixed-mode surveys.

• Face-to-face delivery—We mentioned in Chapter 1 (Volume I) that Don Dillman et al. have an excellent history of surveys.¹ According to Dillman, face-to-face or in-person interviewing dominated during the first two-thirds of the 20th century.
• Mailed delivery—Dillman goes on to say that by the “early 1980s the dominant mode of surveying for government surveys approved by the Office of Management and Budget was mail.”²
• Telephone delivery—He notes that “in the early 1980s the telephone . . . almost completely replaced in-person interviews for surveys of the general public.”³
• Web delivery—In the 1990s web surveys became still another mode of delivery and continued to grow in popularity during the first decade of the 21st century.
• New technological advances led to many variations and offshoots of these four basic modes of survey delivery.⁴ Here are just a few examples.
  • CATI—Computer-Assisted Telephone Interviewing, where telephone interviewers sit at a computer workstation and read the questions off the monitor and enter the respondent’s answers on their keyboard.
  • IVR—Interactive Voice Response, where computers are programmed to administer the survey and respondents enter their answers on their touchtone phones.
  • CAPI—Computer-Assisted Personal Interviewing, where interviewers use a computing device, such as a laptop or a tablet, and enter the respondent’s answers directly on the device.
  • CASI—Computer-Assisted Self-Interviewing, where respondents enter the information themselves directly on the computing device.
  • Web surveys using smartphones and tablets.⁵
  • Address-based sampling, which uses the U.S. Postal Service’s Delivery Sequence File to select a sample, has led to an increased use of mailed surveys.⁶

All these advances have led to a veritable explosion of different forms of survey delivery.

Another important change has been the increase in surveys that use more than one mode of delivery, which are often called mixed-mode surveys. As Dillman notes, this has allowed researchers to take advantage of certain features of each mode and to “compensate for the inadequacies” of the different modes.⁷

Survey delivery varies along several important dimensions.

• Some surveys are interviewer-administered while others are self-administered. Face-to-face and telephone interviews are administered by interviewers, while mail and web surveys are typically self-administered. The presence of an interviewer can influence what people tell us or even if they’ll respond.
• These modes of delivery provide the researcher with different amounts of control over the administration of the survey. Mailed surveys provide virtually no control over the administration of the survey, while web, phone, and face-to-face interviewing provide more control. Control over the administration of surveys allows researchers to more easily do things such as randomize the order of questions and response categories.
• Some modes of delivery provide greater personal contact between the interviewer and the respondent. Web and mailed surveys allow no personal contact, while phone surveys provide some contact, and face-to-face interviewing allows the most contact. Personal contact has both advantages and disadvantages.

Our focus in this chapter is going to be on the four basic modes of survey delivery—face-to-face, mail, telephone, and web—and on combinations of these, called mixed-mode surveys.

Face-to-Face Survey Delivery

Face-to-face interviewing is an interviewer-administered survey, which provides maximum personal contact between the interviewer and the respondent and allows the interviewer to control the actual administration of the survey. Face-to-face interviewing dominated during the development of interviewing methods. Many of the early developments of interviewing techniques can be applied to other modes of survey delivery, but we will discuss them under face-to-face interviewing in this section.

Survey research is based on the assumption or goal that all respondents understand each question in the same way. If that is not the case, then respondents would in effect be answering different questions. Nora Schaeffer and Stanley Presser point out that “different groups within the target population may use different vocabularies.”⁸ While we realize that this goal is probably not completely attainable, we need to make every effort to approximate this ideal.

The question, of course, is how to do this. Stephen Richardson et al. suggest two very different approaches.⁹ In a scheduled or, as it is typically called, standardized interview, the interviewer asks the same question using the same question wording and the same question order for all respondents. Think of a medical experiment where we are trying to determine the effect of a particular drug. The drug is the stimulus that has certain effects. In order to determine the effect of the drug, subjects in a particular experimental group must receive the same stimulus. In an interview, the questions and the order in which the questions are asked are the stimuli. So all respondents must receive the same stimuli, which means that the question wording and order must be identical for all respondents.

In nonscheduled interviewing or, as it is typically called, nonstandardized or conversational interviewing, the wording and order of the questions vary from respondent to respondent in order to make sure that all respondents infer the same meaning to these questions. As Richardson puts it, if we want the questions to have the same meaning for each respondent, then “they must be formulated in wording that is appropriate for each respondent.”¹⁰ In other words, the interviewer must vary the wording and sequence of the questions in order to keep the meaning constant.

Schaeffer and Presser point out that

Respondents who are uncertain about the intent of a question may ask the interviewer for clarification. Yet the dictates of standardization (motivated by a concern that all respondents hear the same information) mean interviewers are usually not allowed to provide substantive help. Critics of standardization have pointed to this as a weakness of the traditional approach to survey interviewing.¹¹

These two different types of interviewing can best be thought of as a continuum from the completely standardized interview, at one end, to the completely nonstandardized interview, at the other end. Most surveys fall somewhere between these two end points, mixing elements of both the standardized and nonstandardized interviews.¹²

Patricia Gwartney provides us with “guidelines for asking questions in the standardized interview.” These include the following.

• “read all questions exactly as written”
• “read all questions in order prescribed”
• “never skip a question”
• “read questions in a deliberate manner”
• “remain neutral”
• “keep respondents on task”
• “use positive feedback to guide and reward participants”
• ask sensitive questions “in a normal tone of voice at a normal pace”¹³

Frederick Conrad and Michael Schober make the argument for nonstandardized or conversational interviewing.

Conversational interviewing . . . is designed to make sure that all respondents understand the question the same way—to standardize the meaning of that question—irrespective of who reads it to the respondent. It thus embodies the assumption that simply speaking words does not guarantee that the listener will grasp their intended meaning; speakers and addressees may engage in further dialogue in order to understand each other as well as they need to.¹⁴

Conrad and Schober describe their study that compared standardized and nonstandardized or conversational interviewing. Respondents were first interviewed using a standardized interview. Then they were reinterviewed 1 week later. A random half was reinterviewed using the same standardized interview. The other random half was reinterviewed using conversational interviewing. When Conrad and Schober compared the first and second interviews, they found that more respondents changed their answers in the second conversational interview than in the second standardized interview and that the changes “conformed more closely to official definitions.”¹⁵ In other words, they found that the changes were in the direction of what the researchers meant to be asking.

Let’s look at some examples. Respondents were asked three different types of questions. One set of questions required a numerical answer. For example, some questions dealt with housing, such as “How many bedrooms are there in your home?”¹⁶ Other questions required only yes or no responses. For example, there were questions about purchases, such as “Have you purchased or had expenses for household furniture?”¹⁷ The other set of questions asked them to list the purchases they made.

When respondents in the standardized interview were confused or unsure of how to answer the questions, interviewers responded in a neutral or nondirective manner. For example, they would reread the questions or repeat the response categories. If they were asked what a question meant, they would respond neutrally by saying “whatever it means to you.”¹⁸

Now let’s look at the conversational interviewing approach. One question asked if respondents “purchased or had expenses for inside home maintenance or repair services.”¹⁹ This was not supposed to include repairs that they did themselves but was meant to refer to repairs that they paid someone else to do. Some respondents misunderstood this. In the conversational approach, the interviewer reminded the respondents that this only referred to repairs that they paid someone else to do.

Another question asked, “How many other rooms are there other than bedrooms and bathrooms?”²⁰ In the conversational approach, the interviewer and the respondent talked over the respondent’s answer to see if it met the official definition of a room. For example, one issue that was discussed was whether a living room and a kitchen that were not separated by a partition counted as one room or two rooms. Another issue that came up was how to count basements and attics.

When Conrad and Schober compared the first standardized interview with the second standardized interview, they found that “57% of respondent’s purchases were consistent with the official definitions” in both interviews.²¹ In other words, a little more than half of the time respondents answered the questions about purchases in the way intended by the researchers. As an example of an answer that didn’t conform to the official definitions, respondents often listed paying phone bills as a purchase, but according to the study’s definitions, this was not a purchase. However, when Conrad and Schober compared the first standardized interview with the second conversational reinterview, they found that 95 percent of the answers in the conversational reinterview met the official definitions of a purchase compared to only 57 percent of the answers in the standardized reinterview.

Regardless of whether we are using a standardized or a nonstandardized interview, face-to-face interviewing has several important characteristics.

• Consider the interaction between the interviewer and the respondent during the introductory phase of a survey where the focus is on getting the respondent’s consent to be interviewed. Since there is more physical contact when we interview someone face-to-face, the interviewer can use the respondent’s body language and voice inflections as cues to help convince the respondent to be interviewed. It’s easier to make a personal appeal to a respondent when we interact face-to-face. Using these cues, the interviewer has more opportunity to tailor the request for an interview to particular respondents.
• Interviewer-administered surveys, like face-to-face interviewing, can rely on the interviewer to probe when the respondent’s answers are unclear or inadequate. For example, if a respondent says that the most pressing problem facing their community is crime, the interviewer can ask the respondent to tell me a little more about that. Interviewers can also help clarify questions when respondents tell us that they are confused.
• The interviewer can make use of visual materials. For example, the interviewer can hand the respondent a card that contains the response categories and ask the respondent to choose the best response. Or the interviewer can hand the respondent a map to confirm that the respondent lives in the area of interest.
• Because the interviewer is present, the possibility of measurement error might be increased. It’s possible that the respondent might give the socially desirable response or be less likely to truthfully answer sensitive questions, such as questions about illegal drug use, because of the interviewer’s presence. Respondents might answer differently to white or black, male or female, or younger or older interviewers.
• Face-to-face interviewing is typically the most expensive of all modes of survey delivery because interviewers often have to go to the location of the respondent and because of the cost associated with paying the interviewers.

Mailed Survey Delivery

A mailed survey is a self-administered survey with no personal contact between the interviewer and the respondent, and the interviewer has little control over the administration of the survey.ⁱ There are a number of good references that will help you learn how to do a mailed survey. We listed some of them in the annotated bibliography at the end of this chapter.

Dillman has written extensively on the different modes of survey delivery. In 1978, he proposed the total design method for doing mailed surveys.²² This approach consisted of two parts.

The first is to identify each aspect of the survey process that may affect either the quality or quantity of response and to shape each of them in such a way that the best possible responses are obtained. The second is to organize the survey efforts so that the design intentions are carried out in complete detail.²³

Using social exchange theory, he discusses how to maximize the rewards and minimize the costs to the respondents and how to ensure trust between respondents and interviewers.

Dillman takes the reader through the steps in designing a mailed survey from writing questions to putting the survey together to maximize the response rate. In other words, the total design method gives you step-by-step instructions on putting together a good mailed survey. We’ll talk about writing good questions in the next chapter of this book and about carrying out the survey in Chapter 3 (Volume II).

In Dillman’s 2000, 2009, and 2014 books, the total design method morphed into the tailored design method.²⁴ The focus shifted to reducing all types of survey error—sampling, coverage, nonresponse, and measurement. We discussed these different types of survey error in Chapter 3 (Volume I). Dillman’s emphasis was on tailoring the survey to fit the population to be surveyed and the information that the survey is designed to obtain.

Mailed surveys have certain important characteristics.

• The motivation to respond must be contained in the cover letter that is included with the survey. You need to carefully consider what incentives you could provide to encourage the respondents to complete and return your survey. Incentives could include a prepaid cash incentive or a small gift or you could appeal to the respondents’ desire to be helpful.ⁱⁱ You might contact nonrespondents by phone and e-mail if that information is available to encourage them to respond, but often you don’t have their phone number or e-mail address.
• All directions for completing the survey must be contained in the written information. For example, if you want respondents to skip certain questions based on their answers to previous questions, then that has to be made clear.
• Your strategy for maximizing response to the survey has to be carefully planned out and implemented through a series of mailings that make different appeals and catch the respondent’s attention.
• There’s virtually no personal contact between the researcher and the respondents in a mailed survey. This means that the opportunity to probe and clarify the respondent’s answer is limited. Probes will have to be built into the mailed survey. For example, if we ask whether a person is employed and what his or her job title is, we can include a follow-up question that asks respondents to describe what they do in a typical week. This is basically a probe question since we are asking the respondent to tell us a little more about what he or she does on the job.
• Because the respondent is actually looking at the survey, you have the opportunity to use visuals. You can include graphics, charts, maps, and other types of visual information.
• Since there is no interviewer present, some types of measurement error might be reduced. Respondents might be more willing to truthfully answer sensitive questions and might be less likely to give the socially desirable response. Without an interviewer, we no longer have to be concerned about the effect of the interviewer’s race or gender or age on what respondents tell us.
• Mailed surveys along with web surveys are usually the least expensive of the modes of survey delivery, because the survey can be delivered to the respondents without an interviewer’s presence.

Telephone Survey Delivery

A telephone interview is an interviewer-administered survey with some personal contact between the interviewer and the respondent that allows the interviewer to control the actual administration of the survey. There are a number of good references that will show you how to do a telephone survey. We listed some of them in the annotated bibliography at the end of this chapter.

Let’s look at some of the important characteristics of a telephone survey like we did for face-to-face and mailed surveys.

• Because telephone interviews are interviewer-administered, the interviewer has a lot of flexibility to use what the respondent says as cues to try to convince the respondent to consent to the interview request.
• Telephone surveys often have significant coverage issues. In 2000 about 2.4 percent of households in the United States lacked any type of telephone (either cell or landline), compared to 5.2 percent in 1990.²⁵ While this is certainly a coverage issue, it’s rather small. A much larger coverage issue is raised by the expanded use of cell phones. About 90 percent of adults have a cell phone and about 47 percent of adults live in households that are cell phone only households.²⁶ Some segments are even more likely to have only a cell phone such as “Hispanics, younger adults, those with lower income, and those living in urban areas.”²⁷ Other potential respondents have both a landline and a cell phone, but it’s difficult to reach some of them on their landlines since they rely heavily on their cell phones. Some segments of the population are particularly hard to reach even on their cell phones. Aria Nilson and Ronald Cossman point out that “young adults may move to attend college out-of-state, retaining their old telephone number. If they reside in a wireless-only household, this creates an ‘unreachable’ segment of the population for geographically-based random digit dialing.”²⁸ As a result of these coverage issues, telephone surveys now include cell phone numbers. The Pew Center has included cell phones in its surveys for some time. Recently they announced that they “will increase the percentage of respondents interviewed on cellphones from 65 to 75 percent in most of its 2016 telephone surveys.”²⁹
• As the phone survey is interviewer-administered, the interviewer can probe to clarify the respondent’s answers or to get more information.
• Because telephone surveys are interviewer-administered, this might increase some types of measurement error, such as the tendency of respondents to offer the socially desirable response and to answer sensitive questions in a less-than-truthful manner. Respondents might also respond differently based on the perceived race or age or gender of the interviewer.
• Telephone surveys can’t rely on visuals unless you are able to send copies of the survey to the respondent, which is unlikely. That means that you can’t make use of visual materials as you can in a face-to-face, mailed, and web survey.
• Both face-to-face and telephone surveys are interviewer-administered, but there is a critical difference. Because you are interacting over the phone, certain constraints on a phone survey are introduced. For example, you have to avoid questions with many response categories. By the time you read the fifth or sixth response category, the respondent may have forgotten the first couple of categories. That may result in a tendency to choose categories that occur at the end of the list.
• It’s possible to capture audio recordings of the telephone interview. That’s particularly important for open-ended questions such as “what’s the most pressing problem facing your community today?” With the audio recording, you can transcribe and enter the respondent’s answer into a computer file and analyze it using specialized software such as NVivo.³⁰
• Computer-assisted telephone interviewing software has been available for many years. It allows you to write computer code that will display the questions and response categories on the interviewer’s computer screen. The software allows you to randomize the order of questions and the order of response categories. You can also randomly assign one form of the question to a random portion of respondents and another form to the remaining random portion.
• Phone surveys are usually less expensive than face-to-face interviewing but more expensive than mailed or web surveys. Interviewers can also contact a larger number of respondents in a shorter period of time than in a face-to-face survey.

Web Survey Delivery

Web surveys are self-administered and often provide either no personal contact or very limited contact between the interviewer and the respondentⁱⁱⁱ but allow the interviewer to control the administration of the survey. As we said earlier about mailed and telephone surveys, there are a number of good references that will show you how to do a web survey. We listed some of them in the annotated bibliography at the end of this chapter.

Let’s look at some of the important characteristics of a web survey, like we did for face-to-face, phone, and mailed surveys.

• Since web surveys are self-administered, that means that you won’t be able to use the respondent’s body language and voice inflections to tailor an approach that would encourage him or her to comply with your request for an interview. However, you might have other information about the respondents, such as the department they work in or their staff position in a business that you can use in your appeal for their cooperation.
• Coverage error depends on the population that you want to study. If your population is faculty at a university or employees in a business, you probably can deliver the survey to your sample with a minimum of coverage error. But for populations such as all adults in your community, you have no way to deliver the web survey to those without Internet access. Internet access varies by country³¹ and by demographic variables such as education,³² although Anja Mohorko et al. report that the digital gap between those with less and more education is decreasing over time.³³
• There are ways to decrease item nonresponse by adding statements following a skipped question, encouraging respondents to go back and answer the question. Terek Al Baghal and Peter Lynn refer to these as “motivational statements.”³⁴
• Even though the interview is self-administered, it’s still possible to probe for more information to clarify the respondent’s answer. General probes can be included to inquire why respondents answer the question as they do. Pamela Alreck and Robert Settle discuss interactive or dynamic probes, in which specific questions are asked depending on keywords that occur in the respondent’s answer.³⁵ You can also provide links to information that might help answer questions the respondents have when they try to answer your questions.
• Web surveys might reduce some types of measurement error. For example, research has shown that respondents are more likely to truthfully answer sensitive questions if an interviewer is not present.³⁶ Respondents might also be less likely to give the socially desirable response to certain questions. Without the interviewer’s presence, you don’t have to worry about the effect of the interviewer’s race or age.
• Web surveys allow you to randomize questions and response categories and to use visuals, such as maps or tables or charts, in your questions. These are powerful tools that can be used because of advances in technology.
• Web surveys, along with mailed surveys, are usually the least expensive of the modes of survey delivery because the survey can be delivered to the respondents without the cost of interviewers.

Mixed-Mode Surveys

Dillman reports that one of the changes in survey delivery is the increased use of multiple modes of survey delivery to take into consideration the advantages of each mode.³⁷ This has occurred for a number of reasons.³⁸

• Some delivery modes are less expensive. Mailed and web surveys have lower delivery costs than face-to-face and telephone surveys.
• Different modes have different types of coverage error. Phone surveys are limited to those you can reach by phone. Web surveys are limited to those you can reach over the Internet. By combining various modes, you can take advantage of the different coverages of each method.
• Nonresponse is always a concern with surveys. By using different modes of delivery, you are often able to increase response to your survey.
• Measurement error is another concern. As we noted earlier, research has shown that the presence of interviewers often affects what people are willing to say. By combining mailed and web surveys with other modes, you can decrease this type of measurement error particularly for sensitive questions and questions dealing with socially desirable or undesirable topics.

There are a number of ways in which face-to-face, mailed, telephone, and web surveys can be combined. You could contact respondents in multiple ways. For example, you could contact them by e-mail or by phone or both. You could also provide respondents with multiple ways of responding to a survey, giving them the choice of replying on the web or by mail.³⁹

In a mailed survey, you could provide respondents with a phone number or a web address that they can use to get answers to their questions or express their concerns about issues such as confidentiality. In a telephone survey, you can mail information to respondents that they might request. In a web survey, you can provide a telephone number that respondents can call to indicate their concern about issues raised in the survey.

Let’s look at a couple of actual examples of a mixed-mode survey. The first example is based on an article by Marilyn Worthy and Danielle Mayclin that describes the “Residential Energy Consumption Study (RECS) [which] is a survey of housing units conducted by the U.S. Energy Information Administration to measure energy-related characteristics, consumption, and expenditures in U.S. homes.”⁴⁰ The survey starts with a face-to-face interview of households, which is “followed by the Energy Supplier Survey (ESS), a mandatory survey of the energy suppliers for each household.”⁴¹ In the past, a mailed survey was used to collect the information from the suppliers, but recently multiple modes of delivery have been used. Suppliers are initially contacted by phone and then by mail to provide the link to the U.S. Energy Information Administration’s website. Once the supplier accesses the website, he or she is given three different ways of providing the information—“paper forms, online forms, or Excel template.”⁴² After the information is received, it is screened for “missing data, anomalous data (such as outliers or inconsistent patterns), or respondent comments.”⁴³ Data quality was determined by the number of companies that did not respond at all (unit nonresponse), did not answer specific questions (item nonresponse), and by the number of times the Energy Administration had to edit the data.

Data quality varied by the way the energy suppliers chose to provide the information.⁴⁴

• Large suppliers were more likely to use the Excel spreadsheet, and smaller suppliers were more likely to use paper or online forms.
• Some suppliers used other modes of responding, such as “other electronic files and nonstandard printouts.” Item nonresponse was lower “among companies that used the standard modes (paper form, online form and Excel template) than those using nonstandard modes.”⁴⁵
• Data quality was higher for the paper and online forms and lower for the Excel spreadsheet. Data changes were made to 31 percent of the cases using the Excel spreadsheet but only 14 percent of those using online forms.

This example not only shows the utility of mixed-mode surveys but it also shows the importance of thorough editing of the data and the importance of following up to resolve problems.

Another example of a mixed-mode survey is the American Community Survey (ACS). Prior to 2010, the U.S. Census used two forms—a short form and a long form. The ACS, which started in 2005, gradually replaced the long form so that only the short form was used in the 2010 U.S. Census. Currently, the ACS is conducted monthly to provide sample data to make estimates for the U.S. population. With the ACS, small areas require 5 years of data to produce estimates, while larger areas require 1- or 3-year samples. For most households, the initial survey was mailed to households. After repeated mailings, telephone calls were made to the households where phone numbers were available. The final phase of the survey was in-person interviews. This series of different survey delivery modes produced very high response rates, and careful editing of the survey data produced very high-quality data.⁴⁶

Meaning of Survey Questions and Answers

Regardless of the mode by which the survey is delivered, what is essential is determining how respondents interpret the questions and what respondents mean by their answers. Robert Groves et al. suggested several different approaches.⁴⁷

• Focus groups—Form a focus group to talk about the issues you are asking about in your survey. A focus group is a small number of individuals who are part of the population for your survey.
• Cognitive interviews—Select some individuals from your population and administer the survey. As they are taking the survey, ask them to tell you how they interpreted the questions and what they meant by their answers.
• Randomized or split-ballot experiments—Assign different versions of a question to random parts of the sample and compare the results to see the effect of question wording.

Howard Schuman also proposed an easy way to determine what respondents mean by their answers. Ask them why. For example, if respondents are asked what they consider to be the most pressing problem facing their community, follow up their answers with probes, such as “Could you tell me a little more about that” or “Would you explain what you mean by _______?”⁴⁸ The same type of probe could also be used in questions that give respondents a choice of several responses. For example, Schuman cites a question asked of respondents during the Vietnam War. The question was “in view of the developments since we entered the fighting, do you think the United States made a mistake in sending troops to fight in Vietnam?”⁴⁹ Possible answers were yes, we made a mistake and no, we didn’t make a mistake. But what did respondents mean by mistake?⁵⁰ An easy way to find out is to ask them what they meant by their answer. Another possibility is to randomly select respondents for probing, what Schuman calls the random probe.⁵¹ A similar approach is suggested by George Bishop, who proposed asking respondents “to think out loud” or to “talk about” how they arrived at their answers.⁵²

These suggestions illustrate several ways to find out what respondents mean by their answers and how they interpret the question. They can be incorporated into most surveys without much difficulty and will help the researcher interpret their data.

Summary

• Modes of survey delivery include:
  • Face-to-face
  • Mailed
  • Telephone
  • Web
  • Mixed-mode
• Dimensions of survey delivery
  • Interviewer-administered versus self-administered
  • Degree of control over the administration of the survey
  • Degree of personal contact between interviewers and the respondents
• These different modes of survey delivery all have advantages and disadvantages.
• The presence of an interviewer provides greater opportunity for convincing the respondent to comply with our request for an interview by allowing the interviewer to observe body language and vocal inflections.
• There is a greater tendency to give the socially desirable answer in an interviewer-administered interview. Respondents are less likely to respond accurately to sensitive questions with an interviewer present. And the interviewer’s race or gender or age might affect what respondents tell us.
• Visual materials, such as maps or charts or tables, can be used in face-to-face, mailed, and web surveys but typically not in telephone surveys. The interactive nature of web surveys allows display of visual materials tailored to particular respondents.
• Audio recordings can be made of face-to-face and phone surveys to allow for better capture of the respondent’s answers to open-ended questions.
• Technological advances in the administration of phone and web surveys make it easy to randomize the order of questions and response categories.
• Coverage error remains a concern in all types of surveys depending on the mode of survey delivery and the nature of the population being surveyed.
• Mailed surveys, along with web surveys, are usually the least expensive of the modes of survey delivery because the survey can be delivered to the respondents without the cost of having an interviewer.
• Mixed-mode surveys are used to decrease cost, increase response and coverage, and reduce measurement error by taking advantage of the different survey delivery modes.
• There are several ways to try to understand how respondents interpret survey questions and what they mean by their answers. These include approaches such as asking why and thinking out loud.

Annotated Bibliography

Face-to-Face Survey Delivery

• Stephen Richardson et al.’s Interviewing—Its Forms and Functions is a good place to start.⁵³ They discuss both the standardized and nonstandardized interviewing approaches, which they refer to as scheduled and nonscheduled interviewing.
• For further reading, look at Raymond Gorden’s two books—
  Interviewing: Strategy, Techniques and Tactics⁵⁴ and Basic Interviewing Skills.⁵⁵

Mailed Survey Delivery

• Don Dillman’s four books: Mailed and Telephone Surveys—The Total Design Method;⁵⁶ Mail and Internet Surveys—The Tailored Design Method;⁵⁷ Internet, Mail, and Mixed-Mode Surveys—The Tailored Design Method;⁵⁸ and Internet, Phone, Mail, and Mixed-Mode Surveys—The Tailored Design Method⁵⁹ are the places to go to learn about mailed surveys.

Telephone Survey Delivery

• Start with Dillman’s books on telephone interviewing mentioned previously.
• Also look at Patricia Gwartney’s The Telephone Interviewers’ Handbook.⁶⁰

Web Survey Delivery

• Mick Couper’s Designing Effective Web Surveys is an excellent discussion of all aspects of web surveys.⁶¹

Mixed-Mode Surveys

• Dillman et al.’s 2014 book mentioned previously is a great source of information about conducting mixed-mode surveys.

Determining the Meaning of Survey Questions and Answers

• Robert Groves et al.’s Survey Methodology⁶² and Howard Schuman’s Method and Meaning in Polls and Surveys⁶³ will help you figure out ways to better understand what respondents mean when they answer your questions.

ⁱMailed surveys can be combined with other modes of survey delivery. For example, those who don’t return the mailed survey can be contacted personally and encouraged to return it. Respondents can also be given a web address and have the option of either completing and returning the survey through mail or completing it on the web. However, the survey, then, becomes more of a mixed-mode survey, which we will discuss later in this chapter.

ⁱⁱSee Chapter 3 (Volume I) for a discussion of incentives.

ⁱⁱⁱFor example, there are some approaches that allow video interface using web cams.