Opt-in and opt-out sampling
Another particularly ethically contentious sampling strategy is the use of ‘opt-out’ rather than ‘opt-in’ sampling techniques. ‘Opt-out’ samples are those where participants are contacted without volunteering to take part in the research – and excluded only when they say they are unwilling to participate. This approach is seen as problematic by many ethics committees because it undermines the principle that consent should be freely given. For example, the fourth principle in the ESRC Framework for Research Ethics states that:
Research participants must participate in a voluntary way, free from any coercion.
(see our Ethics Principles page to consider what this means in relation to different aspects of your research)
With opt-out methods, participants may be included unless they actively say ‘no’. But people may find it difficult to say ‘no’ to a researcher for a variety of reasons. Not least, for fear of causing themselves problems in relation to the service or organisation through which they have been recruited.
For example, imagine you have applied for residential respite care for an elderly relative, and you get a call from a researcher who says that their social worker has given permission to ask you to be in their study. Can you say no? Will the researcher tell your social worker that you didn’t help? Will your social worker think you are being uncooperative? Could it affect their decisions about your case?
For example, imagine you are a child in school and a researcher comes to your class. They want to ask about children about their daily lives, and they ask people to put their hands up if they don’t want to take part. Your parents are in the process of separating, and you don’t want to talk about it, but if you put your hand up, you’re afraid your friends will ask why you didn’t help. How would you feel?
For example, imagine that your boss has been asked to nominate workers to take part in interviews in a management study in your organisation. You do not want to do it - you are feeling very over-worked, and have serious concerns about the management of your workplace, and are worried that your views might be identifiable. But can you say no? Or is your boss entitled to ask you to do this, as part of your professional role?
In addition, practical considerations about how to offer participants the opportunity to opt-out can lead to ethical challenges. By the time a participant is given the opportunity to opt-out, data about them may already have been accessed and processed in a number of ways. Even asking participants to opt-out can seem potentially coercive; opting-out usually involves taking some action (e.g. speaking to someone, or returning a letter) in order to avoid participating in a project. The action may be more or less arduous, but the consequence of not doing it is being included in research that you may not wish to be a part of. This may be experienced as potentially coercive.
The alternative to an opt-out approach is to ask participants actively to ‘opt-in’ to your study – to volunteer to take part. Ethics committees are likely to expect you to prefer opt-in sampling unless you have good reasons that justify an opt-out sampling strategy (go back to the checklist of questions on assessing risk and harm). If you decide to adopt an opt-out approach you need to be prepared to justify it in detail, and very carefully, both in your grant application and later in your application for ethics review. Because a committee may question the use of an opt-out approach, you may also need to allow more time to get through ethics review.
When is ‘opt-out’ sampling justifiable?
This is a hotly contested issue in social science and medical research. There is a range of evidence showing that opt-in samples are less representative than samples recruited by opt-out methods. See our discussion of excluded groups in research, where we noted that sampling strategies may inadvertently exclude certain categories of people.
Opt-in methods result in lower response rates, and have also been shown to bias samples towards populations with fewer ‘problems’ or lower levels of material disadvantage. For example, a study of angina, by Junghans and colleagues (2005) reported that patients recruited by opt-in methods had fewer risk factors for angina, less treatment, and less functional impairment than patients recruited through opt-out sampling - so the opt-in sample was skewed towards patients with fewer of the characteristics that the researchers wanted to study.
Sample bias can produce an incomplete picture and misleading findings. It has been a particular concern for large scale epidemiological research. You can find many examples in the medical literature of concern that scientific rigour – and the quality of the research – suffers as a consequence of requirements for opt-in sampling (e.g. Hewison and Haines 2006). This could also be seen as an ethics consideration - on the grounds that it is unethical to do poor quality research, and potentially it is a waste of public money (if the research has public funding).
The ESRC Framework for Research Ethics - along with other ethics frameworks and professional guidelines - allows exceptions to the principle of freely given consent, and so allows that opt-out sampling may be justified. However, as an exception to this fundamental ethics principle, you can expect opt-out sampling to be scrutinised very closely by an ethics committee.
Consider the following:
- How important is it that your sample is representative of the population you are studying? This depends on your design. For example, in a large-scale normative or epidemiological study, representativeness is likely to be critically important. Representative sampling is less important if you are doing an in-depth qualitative study of a small number of individuals, or if you are using other techniques, like purposive sampling or case-matching, to target participants with particular characteristics. Are you studying a topic where opt-in sampling is likely to skew your data significantly?
- How much does your response rate matter for the methods you plan to use? For example, are you using standardised measures where the norms and interpretation of data depend on you achieving a certain response rate? Are you likely to achieve that with an opt-in method?
If you are considering an opt-out approach, you need to distinguish between:
- using opt-out in relation to the initial approach to potential participants; and
- opt-out in consent itself.
A further complicating issue is whether the participants themselves are opting in or out, or whether a gatekeeper (such as an adult) is being asked to consent on their behalf. Consider the following examples:
A research team are studying young people’s views of sex and relationships education in school. They are conducting a school-based national survey, and are aiming to identify sources of variation in terms of factors including school- and family-level indicators of deprivation, child age, ethnicity and religious background. The researchers know that if parents are asked to give opt-in consent, they will get a smaller, and skewed sample. They propose that: (a) parents are sent a letter and information sheet about the study to inform them that it is taking place, and to give them the opportunity to opt their child out of the survey; (b) young people are given the questionnaire in school, by a member of the research team, who provides an information sheet and explains to the young people that it is entirely up to them whether or not they complete the survey - it is not a school requirement, and that they can return a blank survey should they prefer. In this case, an opt-out approach is likely to be justifiable, providing that adequate steps are taken to ensure that the participants themselves can freely consent - or decline - to take part in the survey.
A researcher is studying good practice in home support services for disabled adults, within one local authority. She plans to write to all adults who are registered with the service, to say that she may get in touch for a phone interview. Her plan is to use consecutive sampling, approaching all those she has contacted (according to a sampling grid based on participant age, gender and disability) until she captures the sample she needs. Is this opt-out approach justifiable? Certainly, potential participants could refuse to speak to her when she calls, but it may be difficult for them to do so. They may not have received, or understood, her letter. This is a study of good (not of representative) practice; the researcher is targeting participants with particular characteristics, and she is working within one relatively small and defined service. So, she could identify a potential sample using opt-in methods. For example, she could ask the service to send a letter or short questionnaire to all service users, asking for their permission to make contact - so she achieves a volunteer sample (which is less likely to be a problem in a study of good practice). Or she could give her sampling frame to the service manager, they could identify prospective participants, and - with careful instructions from the researcher to ensure that those approached do not feel pressured to take part - the manager (or a worker who is known to the participant) could make initial contact to request permission for a researcher to get in touch. The researcher herself would then seek consent herself with those who agree to be contacted.
With both these examples, the key question is whether the planned approach is justifiable. In the latter example, the researcher’s opt-out approach might be judged to be justifiable in a study whose objectives depended on representative sampling, especially if it was not feasible to ask service managers to secure initial permission to contact. But they are not really justifiable for the study proposed.
In general, if you are proposing an opt-out approach, consider two points:
(1) You can expect close scrutiny and challenge from an ethics committee, so take particular care to explain and justify your approach in your application. They will ask if your approach is really necessary, so (while it sounds obvious) you need to be sure that it is. See our further reading for supporting literature, as well as our guidance on writing your ethics application.
(2) Make sure you distinguish between opt-out in approaching your sample and opt-out in consent itself. Even if your participants are contacted through opt-out sampling methods, they should normally be able to give their active consent - or not - to take part in your research. Very exceptionally, ethics committees will approve research designs that do not permit active consent, but this is usually in studies where it is not possible - such as observation of crowd behaviour in a public place. See our section on covert or deceptive research.