Nicoletti et al. (2001) consider the effects of non-sampling errors on the quality of
various income measures in the ECHP. They consider both the effect of unit non-
response, distinguishing between attrition and new/re-entry participants, as well as
item non-response, distinguishing between those who provide partial information and
those who do not respond to any income questions. They find that for those
households with both unit and partial item non-responses final imputed household
income is higher than for responding households. In addition, for those with
complete income non-response final imputed household income is lower than for
responders. They also find that full income non-response is common for those who
are self-employed.
Essig et al. (2003) use a controlled field experiment to consider interviewer and mode
(personal interview and drop off questionnaire) effects on item non-response to
income and financial questions. They find that whilst the mode of the survey affected
non-response, that is the interview had a higher response rate than the drop off
question, there was no additional effect on the respondents’ propensity to respond to
financial questions. They also found that respondent, household and interviewer
characteristics do not have a strong and consistent effect on item non-response to
income questions.
Lynn et al. (2004) used a sample of low income respondents from the ECHP to
consider the effects of interview style on income response. They designed an
experiment to compare dependent interviewing (both proactive and reactive) with
traditional independent interviewing. Dependent interviewing was found to have less
non-response for income than independent interviewing, especially for income
sources which are relatively common or easy to forget.
Schrapler (2003) uses the British Household Panel Study (BHPS) to consider income
non-response in panel studies. He finds that refusals and don’t knows to gross
income relate to different characteristics of respondents. Those who refuse are
mainly male without dependent children whilst those who don’t know are mainly
females, in low or middle occupational groups and work irregularly. Interviewer/area
effects are also found.
Riphahn et al. (2002) use the German Socioeconomic Panel (GSOEP) to consider
item non-response on income and wealth questions. They find that if an interviewer
is female, especially if the respondent is female, there is a higher non-response rate
on income. In addition if the respondent is younger than the interviewer this
improves the response to income questions. They also find that don’t knows are
different in their characteristics to other non-responders in the sample.