Conferences

The elderly population is growing consistently across most European countries. Usually, standard procedures derived from survey methodology are adapted for surveys in the elderly population. However, several studies have indicated that older respondents are prone to survey error to a larger degree. For example, item non-response, non-differentiation, extremity bias, and other indicators are more pronounced among the elderly compared to younger adults.

In the literature various causes for the reduced data quality of responses obtained from the elderly are discussed. Among others, the respondent’s cognitive resources are held responsible for item non-response. Since older respondents are assumed to have a reduced working memory capacity as well as other cognitive limitations they are expected to have more pronounced problems with answering survey questions. In light of these studies, the problems are especially prevalent in behavioral questions and factual questions. By contrast, other scholars assume that older respondents have higher expectations regarding the quality of their behavioral and factual reports which in turn reduces their willingness to provide unreliable responses. Thus, they assume that higher levels of item non-response among older respondents do not necessarily indicate poor data quality. Instead, item non-response should be treated as sign of optimizing response behavior.

In this Paper we will focus on the item non-response in the data obtained from elderly respondents in the first wave of the SHARE study. We will test whether the occurrence of item non-response can be correlated to the available measure of the respondents’ cognitive resources. We will present findings from a multilevel logistic regression (STATA gllamm) assessing the driving factors for the occurrence of item non-response. In addition to question characteristics and socio-demographic characteristics of the respondent we will asses the impact of the respondent’s cognitive resources.

Results confirm the hypothesis according to which question properties as well as the socio-demographic characteristics of the respondents have a significant impact on the occurrence of item non-response. In addition, a reduced working memory capacities leads to more item non-response. Similar effects are observed for measures of reasoning and fluency. Findings are discussed in light of the contradictious explanations of item non-response among the elderly.