3. Research hypotheses
Natalja Menold
If interviewer impact in terms of substitutions is operationalized using Sodeur’s method, it is expected that it can be observed in survey statistics as deviations from the 50/50 gender ratio in subsamples with respondents as representatives of heterosexual couples. This interviewer impact is expected to differ in terms of varying contactability or cooperation on the part of the partners. Partners differ in terms of contactability in households comprised of couples with young children in which men are the breadwinners (Hoffmeyer-Zlotnik 2006; Sodeur 2007; Stoop 2004). If interviewers apply substitutions the proportion of men should be significantly lower than the true value (50%) in such households, since men are more difficult to contact than women. This pattern changes when taking retired partners into consideration. Here, as previously discussed by Hoffmeyer-Zlotnik (2006), partners have comparable levels of contactability but they may differ in terms of their cooperation. For subsamples of retired couples, the proportion of males is expected to be significantly higher than 50% in the case of interviewer impact due to substitutions. The hypothesis describing this impact in different types of households is:
Hypothesis 1 (H1): Deviations from the true gender ratio (50/50) vary depending on the type of household. In households comprised of couples with (young) children the proportion of males is lower than 50%, while in households comprised of retired partners this proportion is higher than 50%.
As was shown in section 2, different sampling methods can be associated with varying degrees of freedom on the part of interviewers to substitute (figure 2.2). Therefore, the following differences between the sampling methods are expected:
Hypothesis 2 (H2): Deviations from the true gender ratio (50/50) vary depending on the sampling method used in a survey. They are lowest with PRS and highest with NRS samples.
If deviations from the population parameter are caused by interviewers deviating from prescribed standards, then they should vary with the sampling method used or with the type of household, which in turn is associated with varying levels of contactability or cooperation on the part of the partners. The deviations should be stable across time when keeping sampling methods constant. However the deviations can correlate with interviewer payment and control procedures, or with the data collectors, who are expected to differ in terms of practices related to interviewers’ work motivation.
Hypothesis 3 (H3): Except for changes in sampling method in a country, deviations from the 50/50 gender ratio are independent of the influence of other changes over time. Thus, they do not vary across different survey rounds. However, interviewer payment, control procedures and change of data collector are expected to correlate with deviations from the 50/50 gender ratio.
- Date modified: