Non-response reduction in DHS surveys: Follow-up fieldwork in Turkey-DHS 2013
Tugba Adali (Hacettepe University)
Melike Sarac (Hacettepe University)
Alanur Cavlin (Hacettepe University)
Ahmet Sinan Turkyilmaz (Hacettepe University)
Keywords: Methodological challenges and improvements, including in the areas of sampling, measurement, survey design and survey response or non-responseAbstract
Despite decreasing non-response trends for face-to-face surveys around the world, Turkey still has very high response rates in household surveys. The response rate for household interviews is 93.3 percent while the response rate for women interviews is 83.9 percent in TDHS-2013 according to DHS response rate calculations. Tremendous effort has been spent to keep the response rates as high as they are in order to minimize non-response and sampling errors. In addition to re-visits, interviewer-switching, having interviewers equipped with official letters, providing respondents with phone numbers of the survey sponsor, provision of intensive interviewer training of 3 weeks (in-class and practical) which included strategies for doorstep interaction; listing operation; and a follow-up fieldwork. The follow-up fieldwork was designed and conducted soon after the main fieldwork was complete and included visits to households that did not respond due to the following causes: Not at home, not at home during survey date, dwelling not found, refused and other. The fieldwork was conducted in 49 provinces where response rates were lower than expected (approximately below 80 percent for households) and could potentially undermine the precision of certain core demographic indicators. In total, 181 clusters (out of all 634 clusters in the TDHS-2013) were visited by 6 field teams of 4 members each. The follow-up fieldwork took approximately 50 days by 24 interviewers and the number of completed household and women interviews increased remarkably.
From a total survey error perspective, this study aims to answer the following question: How much did the extra fieldwork help in terms of reducing non-response error? In order to investigate this, we looked at 1) how many interviews we gained from the process and 2) how the characteristics of households and women were different for those obtained from the main fieldwork and the follow-up fieldwork. We analyzed the household and women level data for the clusters that were included in the follow-up fieldwork conducted in 49 provinces in Turkey. We computed some basic statistics, and our preliminary findings showed that the number of dejure and defacto HH members; HH wealth; proportion of single-adult HHs and the proportion of HHs with children under 5 are significantly different for the main and follow-up fieldwork. The analyses of the characteristics of women showed some attitude questions to differ, rather than socio-demographic indicators.
On the next step, we will compute estimates of non-response error for our selected statistics on the HH and women level. We recognize the high cost associated with the follow-up fieldwork and the gains in precision of the estimates, and through this study we will be able to conclude on whether or not it was significant in decreasing potential non-response bias. Our preliminary conclusion is that the follow-up fieldwork was worth the effort and is a good strategy in Turkey in terms of increasing response rates, decreasing non-response error, and finally contributing to the