A Fixed Effects Approach to Assessing Bias in Proxy Reports

Abstract

Researchers often have to rely on one person to inform them about the characteristics and behaviors of another person. Recent efforts to study autobiographical memory, however, indicate that such proxy respondents encode, store, retrieve, and report memories differently than respondents reporting about themselves. These differences may make reports by proxy respondents more susceptible to bias. Unfortunately, many of the same practical challenges that make investigators reliant on proxy respondents also make it difficult to collect the validation information necessary to identify sources of bias in these reports. This study proposes a simple and cost-effective procedure for assessing certain common sources of response bias in proxy respondents’ reports. This procedure is based on the idea that members of the same group frequently can be presumed to share the same value on important factual or behavioral survey items. When this is the case, these methods can be used to identify sources of bias in proxy respondents’ reports as long as information from more than one group member has been collected for some groups.