# Full text: Wehner, Sigrid: Exploring trends and patterns of nonresponse

```3. PREDICTING NONRESPONSE
59
essential meaning of the job variable is retained: this states that persons with fewer jobs are
more likely to be nonrespondents.
The interview method "capi" is a highly significant predictor for the nonresponse
study. This seems to assist the "refusing generation" theory, as the general availability of the
telephone can be assumed to be equal within the cohort. A special selection mechanism might
instead be the explanation: the capi method is the reason for the successful conversion of soft
refusals.
The calculated predicted probabilities for this model are shown in the next table.
Predicted Probabilities for Nonresponse in Cohort 1930 / Wave 2 1996
TABLE 10:

Cohort 19301 Wave 2 1996

predicted probabitiy for
in SEDor block party
no party manbershp
keeping number of jobs up
Kelhood of being
until 1991 at njob91-5
MEN
MOME
WOME
in group
MEN
a nonrespondenti
0,031
0.014
0,018
lowest
cati
married 1996 and
interview method is:
0.024
0,159
0,098
0,077
0,126
capi
0,080
cati
interview method is:
married 1996 and
0,048
0,062
0,037
0,339
0,185
0,229
0.281
capl
0,048
0,061
cati
0,079
hterview method is:
0,107
unmarried 1996 and
0,227
0,398
0,335
0,278
capi
cat
0.234
hterview method is:
0,189
0,119
0,151
unmarried 1996 and
capi
0,642
0,510
0,577
0,442
highest
mean value for tjobs in
p20,10 i bold
complete group is 5,3
(Predictions were also calculated if combination of values was not observed.)
To keep the table clear, the number of jobs up until 1991 is kept at 5, which is approximately
the average number (5,3). The table reveals the great influence of the capi variable. Given a
personal interview, the probability of being in the NRS file is almost four times as high as for
telephone interviews (e.g. "men, unmarried, no party, grade 10" p-0,119 for cati, but p-0,442
for capi).
Summary for Cohort 1930:
Being unmarried and having stable careers explain the higher risk of being in the nonresponse
study. Participation in political parties is not significant, whereas the most predictive power
lies in the personal interview method. This combination of attributes can be interpreted as a
weak confirmation of the refusing theory, since particularly the capi-variable seems to be
related to the refusal rate. As the general predictive power of the model is rather weak,
however, we have no strong forecast for nonresponse.
```

## 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.