They report that wealthier households, smaller households, and households headed by women were sometimes more likely to respond, but there are substantive differences by round and by region.5 Response Rates and Attrition Rates To complement the information already published by the World Bank about phone survey attrition, we analyzed data from surveys in two of the six countries in the World Bank’s. [...] We find significant differences between the two groups—namely respondents and nonrespondents— for three out of seven CATI rounds of the survey conducted in Senegal and all of the 11 rounds in Tanzania. [...] We compare the baseline sample to each phone survey round on all of the covariates tested for attrition individually, finding that no characteristics are statistically significantly different from the baseline at p < 0.05 in any round of the two surveys. [...] Figure 2: Composition of the phone samples compared to the face-to-face baseline sample Note: This reports the results of four linear models estimated by OLS, each comparing the average value of one characteristic between the respondents of the different survey rounds. [...] Implications When planning new longitudinal surveys, researchers must anticipate the likely response rates (percentage of the sample frame who complete a given survey) as well as sample attrition (percentage of those enrolled in the study who are lost to follow-up) from different study designs, 4 and gauge the impact of such attrition on the quality of inferences.
- Pages
- 5
- Published in
- United States of America
