Proxy Respondent

If a respondent reports on the properties or activities of another person or group of persons (e.g. an entire household or a company), the respondent is said to be a proxy respondent. In some cases, proxy responses may be a part of the design of the survey. In the U.S. Current Population Survey, for example, a single responsible adult is asked to report for all members of the household 14 years of age or older. In many surveys, adults are asked to report for children. In other cases, proxy responses are used only when there is a particular reason that the targeted person cannot report. In household travel surveys, for example, data are typically sought directly from each adult member of the household. Only if a particular adult is on extended travel, has language or medical difficulties that would interfere with responding, or some similar reason, would a proxy respondent be used.

Since a proxy response is treated the same as a self-reported response, an obvious benefit of allowing proxy responses is to increase the response rate. If the targeted person is unavailable for the entire survey period, the only way to get a response may be to accept a proxy response. In surveys in which information is sought about all members of a household, the lack of a response from one member, if only self-reported data are permitted, could jeopardize the utility of the information from the others.

Not allowing proxy responding also may increase nonresponse bias. Those unavailable to respond for themselves are more likely to be on long trips, in the hospital, or away at college, and so on, than those who are available. If these factors are related to the purpose of the survey, then the use of proxy respondents should be considered.

Because the proxy respondent will have a different perspective and set of memories than the targeted person, proxy responding can be expected to affect measurement error in the survey. Proxy respondents may record fewer less salient events (e.g. smaller purchases or short trips) than respondents reporting for themselves. On the other hand, there are instances when proxy responses may be more accurate than reports from self-respondents, for example, the main record keeper in a household reporting for other household members.

Some survey items do not lend themselves to proxy responding because measurement error is apt to be particularly great. A noteworthy case is attitudinal items. Even if the proxy respondent knows the targeted person extremely well, the attitudes of the proxy respondent will likely be confounded with the attitudes of the targeted person in the responses.