Dropout rates and response times of an occupation search tree in a web survey

Occupation is key in socioeconomic research. As in other survey modes, most web surveys
use an open-ended question for occupation, though the absence of interviewers elicits
unidentifiable or aggregated responses. Unlike other modes, web surveys can use a search tree
with an occupation database. They are hardly ever used, but this may change due to technical
advancements. This article evaluates a three-step search tree with 1,700 occupational titles,
used in the 2010 multilingual WageIndicator web survey for UK, Belgium and Netherlands
(22,990 observations). Dropout rates are high; in Step 1 due to unemployed respondents
judging the question not to be adequate, and in Step 3 due to search tree item length. Median
response times are substantial due to search tree item length, dropout in the next step and
invalid occupations ticked. Overall the validity of the occupation data is rather good, 1.7-7.5%
of the respondents completing the search tree have ticked an invalid occupation.