The WG questions are intended to be administered individually to each person selected to be a respondent in a data collection. In some population censuses or sample surveys, only specific members of the household are selected as a sample respondent while in others all members of the household are included. In the latter case, the questions may be administered to a household respondent who responds for all members in the family or household rather than asking each member to respond for him or herself. Relying on a household respondent administration style of data collection may require question wording adaptations. Questions that are part of the Washington Group tools, which were designed to be asked of each person, can be modified so they can be asked of the household respondent.
For example, the following WG question is asked directly of the subject:
- Do you have difficulty seeing even if wearing glasses? Would you say no difficulty, some difficulty, a lot of difficulty or cannot do at all?
There are two ways of modifying the question when a household respondent is chosen to answer on behalf of each individual in the household:
- Does [NAME] have difficulty seeing even if wearing glasses? Would you say no difficulty, some difficulty, a lot of difficulty or cannot do at all?
Note that [NAME] is determined according to list of names from the family/household roster.
- Does anyone in this household have difficulty seeing even if wearing glasses? Response: Yes/No
- If Yes: Who would that be?
[Individuals determined according to list of names from the family/household roster.]
- Would you say that he/she has no difficulty, some difficulty, a lot of difficulty or cannot do at all?
The administration style of the first alternative is closer to the version used when asking the questions of the survey subject directly. The wording of the question and the response categories is not altered from the original WG question and the question is repeated for each person on the household roster.
The second alternative, while possible and has been used, runs the risk of an unintended loss of information particularly for household members who may have mild difficulty [some difficulty] in a particular domain of functioning. In this instance, the first question functions much like a screening question with a dichotomous yes/no response set. The respondent has not heard the answer categories and may not be considering the full range of functioning when responding to the question. This can be alleviated to some extent by adding a statement to be read before the question is asked that introduces the question and the answer categories. For example, the following could be included – “The next questions ask about difficulties members of your family/household may have doing activities. Consider the full range of difficulties when answering: no difficulty, some difficulty, a lot of difficulty or cannot do the activity at all when answering.”
The WG has not cognitively tested either alternative or the example of text that could be added to Alternative 2, therefore, we have no evidence to support the use of these methods of question administration – or how it may impact the results. However, we are aware that this method is used. Furthermore, evidence from a mixed-administration style methodological test of Alternative 1 in the U.S. National Health Interview Survey, where both approaches (asking each person to report about themselves and Alternative 1) were implemented, suggests that the household administration style using Alternative 1 yields slightly lower estimates of disability than administering the questions to each selected household member directly.
Given the differences that may result from varying question wording across different administration styles, the WG recommends survey planners and managers take into consideration the effects that administration style may have on the resulting data and the estimates produced from these data. As much as possible, varying the wording or number of questions or the response categories should be minimized.