Probability-based versus quota-based surveys? Selection and self-selection biases persist among the oldest old

Abstract

As Jan-Lucas Schanze shows, probability-based surveys better reflect the share of people “85 years and over” in samples than quota surveys. But they still seem to be affected by the same selection/self-selection biases as quota surveys. In particular, they include an excessively high share of voters, even in segments with high abstentionism, such as the “85 years and over” group. They do not enable the detection of the beginning of the decline in participation from the age of 70, or the acceleration of the abstention trend from the age of 80, let alone the collapse of participation among those “85 years and over.” They are no more effective than quota surveys in identifying the evolution of participation with aging. This finding seems to confirm that the issue here is not so much sampling techniques as particularly powerful selection/self-selection biases in the oldest age-group: among those “85 years and over,” individuals who are healthier, cognitively higher functioning, and therefore most active socially and politically most readily agree to participate in surveys. As a result, subsamples of the elderly and very elderly are unrepresentative of the age-groups they are supposed to represent.