AAPOR Guidance for IRBs and Survey Researchers
This document contains the following sections:
Additional topics are addressed in the companion FAQs document.
- Survey Participation: Risk and Benefits
- Informed Consent
- Mandatory Reporting
- Compensation and Incentives
This statement is prepared as an aid to researchers and Institutional Review Boards (IRBs) in their deliberations about whether appropriate consideration has been given to the protection of human subjects when a study involving survey interviews or questionnaires is proposed. Consistent with the recommendations of a 2003 National Research Council report (http://www.nap.edu/openbook.php?isbn=0309088526), we encourage IRBs to provide a review process that is commensurate with the level of risk associated with the specific research design proposed. The guidance is based on extensive experience with surveys by the members of our association and is intended to address a wide range of situations. Since it cannot address the particulars of every case, AAPOR, through its Standards Committee, would be glad to assist you if asked about a specific research proposal. Because the Federal regulations specifically require that IRB members possess expertise or consult with experts in the areas of research that they review, we invite you to contact us to assist with reviews if survey research expertise is not available on your local IRB or in your institution.
Human subjects protection is an ever-evolving topic, so AAPOR updates this document periodically to reflect new knowledge, regulations, and best practices. We encourage researchers and IRB members to stay up-to-date on major developments in the regulatory environment, such as proposed revisions to the Common Rule (National Research Council 2014).
II.a. Assessing Risk
- Survey Participation: Risk and Benefits
Participation in surveys rarely puts respondents at more than the minimal risks of everyday life. This fact is recognized explicitly in the Federal regulations which list surveys as examples of research that may be exempted by the IRB or handled with an expedited review process. Unless the intended participants are minors, in many cases, surveys meet the requirements for exemption provided in 45 CFR 46.101b and presented in the category description below.
Research involving the use of educational tests (cognitive, diagnostic, aptitude, achievement), survey procedures, interview procedures or observation of public behavior, unless:
(i) information obtained is recorded in such a manner that human participants can be identified, directly or through identifiers linked to the participants; and
(ii) any disclosure of the human participants' responses outside the research could reasonably place the participants at risk of criminal or civil liability or be damaging to the participants' financial standing, employability, or reputation. [Emphasis added]
Participation in certain surveys can, however, put respondents at significant risk when, for example, the inquiry concerns stigmatizing or illegal activity and inadequate attention is paid to ensuring respondent anonymity or the confidentiality of responses. We discuss various questions about surveys below to highlight this distinction.
1. Does simply asking a question cause some respondents mental distress? Our experience indicates that some questions may cause momentary unhappiness when asked. Unemployed persons may become upset in talking about being fired or out of work. People may become unhappy talking about their own illness or that of a friend or relative. However, these risks are well within the range of day-to-day experiences and activities. There is no known case of any person sustaining any lasting physical or psychological harm from a survey interview. On the contrary, many people report that talking about themselves and their problems is therapeutic. Analysis of response rates across studies indicate that people are more willing to be interviewed about their work or health, and to spend a longer time giving interviewers what might be considered "personal information" than they are to talk about public affairs or foreign policy.
2. Does answering a question cause some respondents mental distress? Respondents have significant defenses against becoming embarrassed by answering a survey question. They can decline to answer if they wish to do so. There is substantial evidence that respondents do choose not to answer when questions that appear to be threatening are asked of them. Interviewers -- strangers to the respondent -- are trained to remain neutral and to treat each answer as legitimate no matter what is said. It is still possible, however, that some respondents who choose to answer a threatening question may feel temporary embarrassment. There is, again, no known case of any lasting physical or psychological harm from participating in a survey interview. There are sensitive topics however, such as questions about disaster survival, suicide ideation, or domestic violence victimization that may cause respondents more than temporary distress. In these cases the researcher must ask whether the benefits of the research outweigh the risks and should have resources available to offer the respondent in the possible event of extreme distress such as hot line number referrals or on-call social workers or other qualified counselors.
3. Are there risks to respondents who admit to illegal or stigmatizing behavior? On some occasions, researchers must ask about matters such as tax evasion, illegal drug use, domestic abuse or HIV-related behaviors. Such surveys can pose significant risks to respondents since they may suffer adverse consequences if individual identities and responses are disclosed. In all surveys, but particularly in these atypical cases, two issues need to be examined by the researcher and the IRB:
1. Have sufficient steps been taken to protect the identity of respondents and the confidentiality of their answers?
2. Do the anticipated benefits to society outweigh the risks?
The ability of IRBs to assess the risks involved in a survey is bolstered when there is mutual understanding and trust with the researcher. We encourage researchers to work with IRBs to develop procedures to expedite implementation of non-substantive changes that do not increase risk to respondents. Minor modifications to protocols such as deleting a survey question or grammatical changes to wording that do not increase the level of risk to the respondent could conceivably be allowed with appropriate notification after the fact if the IRB in question knows the researcher is familiar with the Common Rule and keeps the rights and welfare of the respondent foremost in his or her mind.
II.b. Anonymity and Confidentiality
It is essential in all surveys that the researcher separate all personally identifiable information (“PII”) (e.g., name, date of birth, home address, email address, telephone number, if known) from the interview itself as soon as it is no longer needed, typically, immediately after the survey has been conducted or the interview has been verified. (Verification is a standard quality control procedure following face-to-face interviewing and is normally conducted soon after the initial interview is completed.) The identifying information (if it needs to be retained) must be stored separately from the interview data in secure files. The level of security needed should be determined according to the level of risk assumed by the respondent and the likelihood that efforts will be made (by, say, law enforcement personnel) to access the identifying information. In cases where data are linked over time, for example in panel studies, identifiers may need to remain with the data for longer periods. In these cases, data security and storage must be a priority.
Additional special measures should be considered to ensure confidentiality of responses in the rare cases where surveys pose significant risks to respondents, or when respondents have been promised confidentiality (see “How much should survey respondents be told?” below). In these cases, the researcher should review alternative methods to ensure that respondents cannot be identified through analysis of the data file alone. The methods may involve adding random noise to the file, suppressing variables, altering demographic information according to a coded system, top-coding/bottom-coding responses or some other procedure. For ensuring confidentiality in published tabular data, additional methods such as controlled rounding as described by Cox (1987) should be considered. AAPOR can recommend consulting support for researchers and IRBs to decide on appropriate methods for treating data files and tabular data in these special survey situations.
For surveys conducted on paper, researchers should ensure that PII is kept separate from the survey responses. In the case of computer-assisted data collection (whether using desktop or laptop computers, or smartphones or other mobile devices) researchers should employ appropriate encryption and other measures to secure the data on these devices. Researchers should also consult their IT staff for assistance in applying data security steps.
Electronic data collection via the Internet, email and SMS has increased in recent years and also presents special challenges for data confidentiality. When any PII or personal health information (“PHI”) is being collected in Internet-based tools, researchers and IRBs should make sure that appropriate security measures are in place to safeguard both the transmission and storage of survey responses. Such safeguards might include encryption during transmission, storage of data on secure servers, and provision of firewalls to protect data from unauthorized access. Special considerations are necessary when PII is collected using hosted services, where the data reside on a vendor’s servers, outside the control of the researchers. For more on special issues related to Internet research, see Kraut et al. (2004) and Singer and Couper (2011).
Electronic data collection through social media platforms such as Facebook, LinkedIn, Twitter, etc., poses special problems related to privacy of those persons connected to survey respondents through the social media who have not explicitly consented to participate in the study. Researchers should consider the types of PII and confidential data that will be gathered on third parties via the consenting participant’s social networks and whether third party consent is recommended or required by their IRB. Similarly, placement of computer cookies in respondents’ computer environment or in their third party social media connections without explicit consent should be assessed relative to written informed consent requirements by their IRB. If collecting data from a social media site, it is recommended that a qualified expert carefully review and advise on the terms of service (TOS) for any restrictions the site may have for researchers.
The availability of geocoded respondent information using geographic information systems (GIS) presents new challenges with respect to data disclosure. Researchers need to assess if tabular data or maps contain enough information to identify an individual participant in a study. For example, if a map showed respondents in a specific housing unit or on a block with enough other information to determine their identity, such as being the only household on a given block, IRBs would consider them to be a potential disclosure risk. Similarly, care needs to be taken with tabular information containing longitude/latitude coordinates or geographic codes at detailed levels that may increase the risk of disclosure.
II.c. Benefits of Surveys to Respondents and Others
Many survey participants report that they enjoy the survey process. This enjoyment and the sense of good feeling they get from helping the research enterprise make surveys possible. The pleasure is probably temporary; no systematic evidence of long-term benefits from survey participation has been collected, though such benefits are possible. (The tangible benefit of payments made to respondents to compensate them for their participation or as a token of appreciation are not considered by IRBs to be a benefit of the research. This topic is discussed at length in Section V below.)
The most obvious benefits of surveys are those to the researcher and to society. Surveys are powerful tools capable of providing generalizable information on a variety of aspects of the human condition when designed and fielded using sound, rigorous methods. Survey data are essential to advancing our understanding of health and disease, explanations of social, psychological, economic and political processes, and evaluation and improvement of public policy. Even where the benefits of surveys are not immediately apparent, the potential benefits clearly outweigh the minimal risk of harm to respondents in the majority of surveys.
In the case of surveys that do pose substantial risks for respondents, the societal benefits are usually, if anything, more clear cut. Studies of HIV-related behavior, for example, are conducted so that better methods of disease tracking and prevention can be developed. These studies are conducted with appropriate safeguards for anonymity and confidentiality to minimize respondent risk, which is potentially very great. When the risks to respondents are not appropriately considered in a survey design, however, the study should not be conducted.
In virtually all survey-based studies, the key elements of consent can be provided to respondents in a concise way at the beginning of a survey in the brief introductory statements of a telephone interview, in a cover letter for a self-administered survey or in the introductory screen in a web survey. This is true regardless of level of risk, and is consistent with the contemporary view of consent as an ongoing process rather than a document. The main elements of consent are: an explanation of the purpose(s) of the study, the approximate amount of time it will take, a description of what the respondents will be asked to do, a description of any foreseeable risks or discomforts, a description of any benefits to the respondents or others, a statement describing the confidentiality of responses, and a statement of the voluntary nature of participation. In face-to-face and telephone surveys, contact information should also be available upon request for questions about the research and about respondent rights. In self-administered and electronic modes, this information can be included in the written introductory information.
- Informed Consent
It is useful to discuss informed consent in surveys using the questions: Who can give consent? How does a survey respondent indicate consent? How much information about the study must be provided to a survey respondent, and when? Included in this discussion should be consideration of the language, culture and literacy level of the target population.
III.a. Who Can Give Consent to Participate in a Survey?
As a general principle, consent can be obtained from adult respondents who can understand the benefits and risks of the survey. Except in the special cases where parental consent itself could pose risks (e.g., studies of child abuse), parental consent must be obtained prior to administration of a survey to a minor and assent (agreement to participate) should be obtained from the child or teen. The age of majority at which a person is considered old enough to legally make their own decisions varies slightly from state to state in the U.S., and from country to country. Researchers should be aware of the age of majority in the geographic region(s) where they are collecting data.
Special consideration must be given to studies being conducted by someone with authority over potential respondents (e.g., teachers, employers, physicians). In these cases, it is particularly important that the respondent recruitment procedures evidence no coercion, either explicit or implicit. Further, it is essential that the researcher make a clear demarcation between research questions and issues arising out of the authority relationship. When the person with authority is a medical provider, there must also be a clear distinction between any medical care being provided for the research and any medical care that is recommended as a form of treatment. As a general rule, we discourage persons with authority over potential respondents from recruiting participants or administering surveys themselves if they will have knowledge about who did and did not participate or have access to individual responses.
III.b. How Does a Survey Respondent Indicate Consent?
In most surveys, respondents indicate their consent by agreeing to participate at the beginning of the interview, and/or by answering questions as they are asked or that appear on a paper or electronic questionnaire. Thus, people may consent to all of an interview, to part of it, or to none of it, depending on how they respond to requests from the interviewer. We should note that significant numbers of people who are approached to participate in surveys refuse to do so or refuse to answer individual questions. There is no evidence that people feel coerced to participate in survey research. Unlike much medical research, which requires "all or nothing" cooperation, surveys permit respondents to opt out easily of parts of the measurement process if they so desire.
Federal regulations (CFR 46.117c) on human subjects protections recognize that written consent forms are not necessary or desirable in every research setting. The regulations provide that, while written consent is the norm in much research involving humans, IRBs may waive requirements for documentation of informed consent if they find that the research presents no more than minimal risk of harm to subjects and involves no procedures for which written consent is normally required outside of the research context. When the written consent form is the only source of identifying someone as a research participant, documentation of informed consent may also be waived.
These conditions describe the vast majority of surveys. As noted above, few surveys pose risks greater than those that the respondent would confront in everyday life. Further, in many contexts, signed consent forms may increase risks to respondents if confidentiality is breached (e.g., research on illegal behavior, health, immigrants, etc.) and reduce cooperation unnecessarily. For example, Singer (1978) conducted an experiment showing the need for a signed consent form reduced response rates by six percentage points. In later research, Singer (2003:280-81) reported that 13 percent of respondents who said they would be willing to take part in a hypothetical survey would be unwilling to sign the consent form. Nonresponse is a primary source of error in surveys, and we attempt to keep it at a minimum so that our samples will be representative of the populations of interest. Signed consent forms, in the case of the typical survey, can increase nonresponse error and increase the burden on respondents without the gain of protecting respondents from significant risks.
Moreover, beyond the possible negative effects of requiring written consent, telephone surveys using random-digit dialing – a common approach in survey research – cannot incorporate signed consent in the protocol prior to the initial contact since respondents' names and addresses are unknown to the researcher. Requiring signed consent prior to the beginning of an interview would, for all practical purposes, make many telephone surveys impossible. Thus, a waiver of documentation of consent is typically the most desirable approach for survey protocols – especially those utilizing telephone and electronic modes.
Even in that small percentage of surveys that do pose more than minimal risks for respondents, the Federal regulations (CFR 46.117c) allow IRBs to waive requirements for signed consent if they find that the proposed research cannot practicably be carried out without the waiver. These types of surveys require a more extensive discussion of risks and benefits with the respondents than do the usual minimal risk surveys. We encourage IRBs to work with the researcher and with the AAPOR Standards Committee to identify means of providing an adequate informed consent process other than the signed consent form when that method would threaten the viability of the research.
III.c. How Much Should Survey Respondents Be Told, and When?
We firmly believe that potential respondents should understand the risks and benefits of surveys, particularly in those rare cases where surveys do, in fact, pose more than minimal risks. In order that the goal of informing respondents not interfere with sound research practice, we must use information methods that do not unduly exacerbate nonresponse bias, jeopardize the measures of knowledge, opinion and behavior in the survey, or induce unnecessary anxiety in respondents.
Researchers and IRBs should consider, in other words, the impact of an informed consent procedure on the objectives of the research. In the normal survey that presents minimal risk, lengthy and detailed information about the objectives of the survey and the questions to be asked is apt to discourage respondent participation and responses without safeguarding respondent rights. In these surveys, the usual practice of a short introduction about the purpose of the study, the approximate amount of time it will take, the sponsor and/or responsible survey organization, and the general topics to be covered is typically deemed sufficient. Respondents should be told their participation is voluntary and they can skip questions they choose not to answer. The goal of the introductory statement and the reminders about the voluntary nature of response is to help ensure respondent autonomy without affecting substantive responses. If appropriate, at the end of the interview, respondents can be debriefed to see if any of the matters covered were upsetting, to give further information on study purposes, or to see if respondents have any further questions.
Information about confidentiality is often included in the survey introduction presented to respondents. This information should be consistent with the procedures actually taken to protect confidentiality and should reflect the degree of confidentiality that can reasonably and legally be promised. More detailed methods of informing respondents may be considered when survey participation does pose substantial risk. The goal here is to provide information necessary for informed choice without unduly increasing nonresponse or response error, which can render survey efforts useless. Each case will have its own unique problems, and so it is difficult to suggest a standard approach. In broad terms, respondents should be informed in an introduction to the survey that the instrument will cover some sensitive topics, but they should not be told so much as to bias their answers (e.g., they should not be informed of the study hypothesis).
This treatment of informed consent in surveys that pose more than minimal risk is necessarily brief given the case-by-case nature of unusual risk. It is important for IRBs to work with the researcher in such cases to craft a procedure that protects respondents from harm while minimizing response and nonresponse bias.
In the course of completing interviews in some studies, respondents may provide information that triggers the mandatory reporting requirements for situations such as child abuse, child neglect, elder abuse or intent to harm oneself or others. If the content or subject of a study is likely to elicit such disclosure and mandatory reporting is compelled, this likelihood must be disclosed as a risk to participants in the consent information. Participants must be informed that there are circumstances where breach of confidentiality is required by law. Reconciling the seemingly disparate responsibilities of confidentiality and mandatory reporting can be a challenging balance. The additional concerns of response and nonresponse bias increase the complexity of this issue.
- Mandatory Reporting
Many professionals are subject to legal requirements to report the intent to harm oneself or others. IRBs generally expect that professionals conducting research will comply with their professional reporting requirements, and although the legal requirements to do so when acting as a researcher may not always be clear, there is an argument that the ethical obligation remains even if no clear legal requirement exists. Similarly, if a study includes testing for medical conditions (beyond collecting self-reported information), such as certain sexually transmitted diseases or tuberculosis, the medical reporting requirements remain in effect and the consent information must indicate the requirement to report the findings to the appropriate state office.
Consent forms should inform participants of these reporting obligations in any study in which reportable information is likely to be evoked. An example of such language is:
There is one exception to confidentiality. If information is revealed about [child abuse or neglect, elder abuse or neglect, or potentially dangerous future behavior to others, etc. as applicable], the law requires that this information be reported to the proper authorities.
There is no need to include this information about reporting requirements if the study design is unlikely to elicit reportable information. It is also important to note that Certificates of Confidentiality (issued by NIH to provide protection to researchers against being compelled in any federal, state, or local civil, criminal, administrative, legislative, or other authority to identify the subjects of research) do not apply to or allow exceptions to mandatory reporting laws. Requirements differ by state and sometimes even county so if an investigator is conducting data collection that is likely to require mandatory reporting, the laws/requirements for all jurisdictions must be reviewed for all locations in order to include the applicable legal risks in the consent documents across locations. Resources for information on reporting laws include:
Department of Health and Human Services Information on Child Abuse and Neglect Reporting
National Conference of State Legislatures Child Abuse and Neglect Reporting State Statue Overview
National Center on Elder Abuse Resources
The practice of offering compensation or financial incentives in survey research has increased considerably over the past decade. Owing to a variety of factors, researchers frequently find it necessary to offer incentives in order to obtain sufficient participation in a study. Incentives may be in the form of cash, gift cards, checks, or non-monetary items (such as t-shirts or other gifts). For research governed by IRB review, researchers need to justify payments based on several variables. There are two key dimensions of concern when designing protocols that include payments to participants. The first is ethical and the second is legal/regulatory.
- Compensation and Incentives
Although payment practices are not directly addressed in the current Federal regulations concerning ethical treatment of research participants, they can become a concern if they are deemed to constitute undue inducement to participate in a study. As outlined in the OHRP IRB Guidebook (1993), the concerns are twofold: 1) would the payments “blind” potential participants to real risks or impair the likelihood that they consider the risks fully, and 2) would the payment induce participants to lie or conceal information that might render them ineligible to participate in the research. With respect to the former issue, research suggests that the typical incentives used in surveys are not coercive (Singer and Bossarte 2006).
The regulatory issue is that there are no specific guidelines for what constitutes an appropriate level of payment or incentive for a given study. IRBs typically look to accepted standards and norms for their community along with specific study factors. A relatively limited commitment of time (e.g., 30 minutes or less) and minimal risk will usually result in approvals of minimal incentive amounts (e.g., <$25) unless the target population is from a high income stratum (e.g., physicians) where higher incentive amounts are common. Small incentive amounts are referred to as “de minimis” payments at some institutions. If the survey data collection is part of a more complex or longitudinal design requiring other more time-consuming practices like diaries or changes in regular routine, higher amounts may be appropriate. As outlined in the OHRP IRB Guidebook, researchers will likely find that views on appropriate payment levels differ across IRBs. Some IRBs may endorse the view that the key aspect of the process is the informed consent – arguing that participants are free to choose to participate as they see fit if the consent is complete and easily understood. Others may be more concerned with ensuring that high levels of payment do not affect the ability of participants to make a choice that is informed and voluntary (that is, that the level of payment is not so high as to result in undue influence). As suggested in the OHRP IRB Guidebook section on incentives, key questions should be considered. The questions most relevant for the majority of studies in survey and public opinion research are:
Are recruitment procedures designed to assure that informed consent is freely given?
What special safeguards are included to protect the rights and welfare of participants who are likely to be vulnerable to coercion or undue influence (e.g., children, prisoners, pregnant women, persons with physical or mental illness, and persons who are economically or educationally disadvantaged)?
Are any incentives likely to unduly influence a prospective participant's decision to participate?
Answering these questions will help researchers successfully justify the use of incentives and compensation in their studies.
Although most survey and public opinion studies fall into the minimal risk category for IRB purposes, we encourage researchers to address all ethical issues relevant to IRB review, including those concerning participant payments. We also encourage researchers to work with their institutional financial officials to assist them in setting reasonable maximums for participant payments and establishing record-keeping policies that balance confidentiality and financial requirements. A thorough examination of any ethical dimensions of the payments along with a review of all applicable institutional policies and guidelines will help ensure that incentives serve to maximize the quality of the research in accordance with all relevant institutional requirements.
Surveys comprise a large portion of research involving human subjects. We in AAPOR believe that it is essential to protect the autonomy of potential respondents and to handle with utmost care the personal information they provide to us. The very existence of the survey as a research tool depends upon a firm foundation of public support. We endorse the aims of human subjects protection regulations and the efforts of institutional review boards to apply them. We want to work with IRBs and researchers to see to it that human subjects protections are applied in a manner that serves the ends for which they were intended. This goal requires distinguishing between surveys and other types of research involving humans, and between the vast majority of surveys that pose little risk and those that pose some risk to respondents. The survey profession has developed methods for ensuring respondent anonymity and confidentiality, and for allowing respondents to express a thoughtful decision to participate or not participate. These procedures differ from those employed in some other forms of research with humans, but they are effective. We urge IRBs to judge how well the survey designs you review embody the principles we have presented here – principles that embrace the fundamental protection process as set forth in the Federal regulations. Finally, AAPOR stands ready to serve as a resource when questions arise about the particulars of any survey design. We can supply advice or consulting support to work out the particulars of difficult survey problems that occasionally arise when balancing good research design and human participant protections.
Duncan, G.T., Elliot, M., and Salazar-González, J.-J. (2011). Statistical Confidentiality; Principles and Practice. New York: Springer.
Cox, L.H. (1987). A Constructive Procedure for Unbiased Controlled Rounding. Journal of the American Statistical Association 82: 520-524.
Internal Revenue Service. Instructions for Form 1099-MISC. Available at http://www.irs.gov/instructions/i1099msc/ar02.html
Kraut, R.M., Olson, J., Banaji, M., Bruckman, A., Cohen, J., and Couper, M.P. (2004). Psychological Research Online: Report of Board of Scientific Affairs’ Advisory Group on the Conduct of Research on the Internet. American Psychologist, 59 (2): 106-117.
National Research Council. (2003). Protecting Participants and Facilitating Social and Behavioral Sciences Research. Panel on Institutional Review Boards, Surveys, and Social Science Research, Constance F. Citro, Daniel R. Ilgen, and Cora B. Marrett, eds. Committee of National Statistics and Board on Behavioral, Cognitive, and Sensory Sciences. Washington, DC: The National Academies Press.
National Research Council. (2014). Proposed Revisions to the Common Rule for the Protection of Human Subjects in the Behavioral and Social Sciences. Committee on Revisions to the Common Rule for the Protection of Human Subjects in Research in the Behavioral and Social Sciences. Board on Behavioral, Cognitive, and Sensory Sciences, Committee on National Statistics, Division of Behavioral and Social Sciences and Education, Board on Health Sciences Policy, Institute of Medicine. Washington, DC: The National Academies Press.
Office for Human Research Protections. (1993). Institutional Review Board Guidebook. Available at
Singer, E. (1978). Informed Consent: Consequences for Response Rate and Response Quality
in Social Surveys. American Sociological Review 43: 144-162.
Singer, E. (2003). Exploring the Meaning of Consent: Participation in Research and
Beliefs about Risks and Benefits. Journal of Official Statistics, 19:273-86.
Singer, E., and Bossarte, R.M. (2006). Incentives for Survey Participation; When are They ‘Coercive’? American Journal of Preventive Medicine, 31 (5) 411-418.
Singer, E., and Couper, M.P. (2011). Ethical Considerations in Web Surveys. In M. Das, P. Ester, and L. Kaczmirek (eds.), Social Research and the Internet. New York: Taylor and Francis, pp. 133-162.
American Association for Public Opinion Research, 2014