Data Quality in Nationwide (German) Social Surveys. Blohm, M. In European Conference on Quality in Survey Statistics, Cardiff, 2006. (ALLBUS)
abstract   bibtex   
"For random probability sampling the use of an appropriate sampling design and a high response rate should ensure high data quality. The following paper investigates the associations between sampling design and data quality on the one hand and response rates and data quality on the other.To answer the first question, different German surveys are studied, which were conducted using different (random) sampling designs. The sampling designs differ in the leeway interviewers have in practice to select the target persons. (Theoretically, in all random sampling designs, the interviewers should have no leeway for this selection when working according to the rules; but in random route designs interviewers select the household and/or the target by themselves, whereas in samples containing the names and addresses of individuals the interviewers have no choice). To get a benchmark for data quality the nation-wide surveys (ALLBUS [1994-2004]) are compared with the German Microcensus. It can be shown, that data quality in sampling designs containing the names of individuals is higher than in other sampling designs where interviewers select the target person and/or the target household by themselves.In a second step, the association between response rates and data quality is investigated over the course of the fielding period. The surveys on the basis of samples with named individuals at different points in time during the fielding period were compared with the German Microcensus with regard to univariate distributions as well as bivariate and multivariate associations. Data quality tends to rise with higher response rates, but after a 30% Response rate is reached improvements in data quality are marginal."
@inproceedings{Blohm2006DatQuality,
  abstract = {"For random probability sampling the use of an appropriate sampling design and a high response rate should ensure high data quality. The following paper investigates the associations between sampling design and data quality on the one hand and response rates and data quality on the other.To answer the first question, different German surveys are studied, which were conducted using different (random) sampling designs. The sampling designs differ in the leeway interviewers have in practice to select the target persons. (Theoretically, in all random sampling designs, the interviewers should have no leeway for this selection when working according to the rules; but in random route designs interviewers select the household and/or the target by themselves, whereas in samples containing the names and addresses of individuals the interviewers have no choice). To get a benchmark for data quality the nation-wide surveys (ALLBUS [1994-2004]) are compared with the German Microcensus. It can be shown, that data quality in sampling designs containing the names of individuals is higher than in other sampling designs where interviewers select the target person and/or the target household by themselves.In a second step, the association between response rates and data quality is investigated over the course of the fielding period. The surveys on the basis of samples with named individuals at different points in time during the fielding period were compared with the German Microcensus with regard to univariate distributions as well as bivariate and multivariate associations. Data quality tends to rise with higher response rates, but after a 30% Response rate is reached improvements in data quality are marginal."},
  added-at = {2016-02-01T15:52:09.000+0100},
  address = {Cardiff},
  author = {Blohm, Michael},
  biburl = {http://www.bibsonomy.org/bibtex/23524f00ee998055d88f3f42d68c97662/gesis_surveydoc},
  booktitle = {European Conference on Quality in Survey Statistics},
  interhash = {d409c4a094a019c85592292aef17d51b},
  intrahash = {3524f00ee998055d88f3f42d68c97662},
  keywords = {ALLBUS 2006 input2014 checked FDZ_ALLBUS inproceedings},
  note = {(ALLBUS)},
  privnote = {ALLBUS_ID=1442 ; Aufgenommen: 22. Fassung, März 2008 ; pubdate=2006/4/24=26.04.2006},
  study = {ALLBUS 1994/ 1996/ 1998/ 2000/ 2002/ 2004/},
  timestamp = {2016-02-01T15:52:09.000+0100},
  title = {Data Quality in Nationwide (German) Social Surveys},
  year = 2006
}

Downloads: 0