jmroemer/CATI_Panel_Attrition

An R analysis of attrition in two mobile phone panel surveys of the World Bank’s Listening to Africa initiative # Attrition in Mobile Phone Panel Surveys Using open data of the World Bank’s Listening to Africa initiative, this project analyzes attrition in two long-term panels that were conducted using computer-assisted telephone interviews (CATI) in sub-Saharan Africa: The Listening to Senegal survey (L2S) performed between 2014 and 2017 and the Sauti za Wananchi survey (SzW) performed in Tanzania from 2012 to 2014. In both projects a face-to-face baseline interview was conducted, and mobile phones were provided to the participants. Over the following two years several rounds of CATI surveys were performed based on the baseline sample. This allows us to analyze potential attrition bias by directly comparing selected characteristics between the group of people who responded to a particular survey round and the ones who did not. In the present version the code analyzes the following characteristics: age, gender identity, whether a participant lives in an urban area, has completed secondary education and whether their main occupation is in agriculture. ## Purpose The R-code provided in this repository was created to analyze the following aspects of attrition in long-term panels: For each survey round it tests weather the group of respondents to this round is equivalent to the group of non-respondents with respect to a set of selected characteristics. If further tests for all CATI rounds and all these characteristics individually if the sample mean of the CATI respondents is equivalent to the baseline sample mean. Graphs showing how response rate, mean of the selected binary variables and age distribution vary over time or survey rounds are created and test-results are saved in csv/html tables. Please refer to the Methods Appendix for further information. ### Outputs * A plot of the response rates of the individual survey rounds against approximate time after the face-to-face baseline. * An attrition table that contains the coefficients of a multivariate linear OLS regressio …