Full text: Brückner, Hannah: Surveys don't lie, people do?

question whether this second aspect of data quality is really different from what I called here 'visible 
data quality. The zero-order correlation is positive but low with r-.0535 but a bivariate statistic may 
be misleading; in any case, it does not prove that they are not generated by the same processes. Are 
the two types of error just two indicators for the same underlying concept (i.e. the propensity to 
produce accurate and complete information)? 
In this case, we would expect effects on data quality to operate the same way for both types 
of errors. Do interviewers who produce on the average many visible errors also produce many 
invisible errors? In section 4.1., I describe differences between interviewers with respect to their mean 
on visible and invisible errors. Finally, in section 4.3.3. I will compare effects of independent variables 
across models. Iwill show differences and similarities in the effect of interviewers, respondents, and 
the interview situation on both types of error. 
Common sense would predict a general decline of errors over field time, since interviewers 
become more acquainted with the questionnaire and more confident in dealing with difficult respon- 
dents. Critics of the survey method might predict a decline in visible errors but an increase in invisible 
errors as interviewers learn to reduce output but hide obvious traces of inadequate behavior. A 
tendency to change question wording with growing experience might contribute to an increase in both 
visible and invisible errors. Table II reports descriptive statistics of dependent and independent 
variables for both error count models.2 The following section describes independent variables and 
hypotheses about their effect on data quality. 
3.4.2. Independent Variables 
The variables described below aim at tapping main concepts identified in previous research 
as determinants of data quality: (1) memory retrieval processes, both in terms of salience and 
remoteness in time; (2) respondents and interviewers ability to deal with the questionnaire; (3) 
potential distractions from the interaction process, both with respect to interviewers work place and 
respondents home; (4) length and volume of the interview as a measure of efficiency or flow' of inter- 
viewing; and (5) interviewer experience and productivity as indicators of organizational dynamics. 
Table II and all the following results are based on 1987 cases. 17 cases were deleted 
because of missing data on the co-variates. I found no indication that these cases differ from the 
included cases in terms of the variables used here.

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.

powered by Goobi viewer