English title: The use of paradata for non-response bias adjustment: A case study on reluctant respondents and fieldwork implementation
Author(s): Hideko Matsuo - Jaak Billiet -
Type: Book chapter
The objective of Chapter 7, by Hideko Matsuo and Jaak Billiet, is comparison of cooperative and reluctant respondents. This paper proposes a new approach for turning reluctant respondents into proxies for all non-respondents using paradata Z-variables (e.g. type of dwelling and neighbourhood). Two types of propensity scores (the focus is on ‘contactability’ and on ‘survey cooperation’) are obtained through logistic regression models and multiplied with each other to obtain non-response weights for reluctant respondents. The effectiveness of these propensity weights is assessed through (1) the level of absolute standardized bias, (2) the level of correlations between Z-variables and R variables (response outcome) and between Z-variables and Y-variables (substantial items), and (3) estimated statistics in substantive models tested between unweighted and weighted respondent samples. Each of these assessments of effect on improvement is also applied to the results of post-stratification weighting based on documented population statistics (age, gender and region). The effects of the two types of weights are different, showing a different measure of bias. The effects of propensity weights are generally weaker than those of post-stratified weights. The paper addresses data quality in paradata and in the fieldwork design on refusal conversion activities.
From page no: 139
To page no: 158
Anthology: Hopes and anxieties. Six waves of the European Social Survey