Page 390 - Special Topic Session (STS) - Volume 4

STS2319 Lakshman N. R. et al.
Mid-term professional 4 5.48 7 6.93 11 6.32
school
University 56 76.7 77 76.24 133 76.44
1
Upper secondary school 13 17.8 15 14.85 28 16.09
1
Average Experience 5.04 6.01 5.60

One of the metrics in the experiment used as a benchmark to assess data
quality is the total number of errors committed by the mode of data collection.
Errors were classified across four categories: skip, validation, logical, and
missing. Skip errors are those where an enumerator failed to implement the
skip rules incorporated into the questionnaire based on responses to
preceding questions. Data validation errors are those that are committed
based on the condition that answer to a question is restricted to certain values.
For example, when a question only expects numeric answers up to 100 but the
enumerators inputs a 200, it will be flagged as a validation error. Logical errors
refer to those committed that fail to meet cross-sectional logic. An example of
this is a respondent answering “Male” to a gender question and “Currently
pregnant” to another question. Finally, missing errors refer to those fields that
were unanswered but were required based on previously set conditions within
the questionnaire.

Empirical Strategy
To investigate the first research question looking at the effect CAPI has on
interview time, the following Ordinary Least Squares model was applied:

= + × + × + × + є

where, refers to the Survey Duration in minutes in the questionnaire of

household i interviewed by enumerator j in cluster c. Meanwhile, = 1 if

the questionnaire was implemented using CAPI and 0 if otherwise. Moreover,
are the household characteristics and are the enumerator
characteristics. The specification also uses enumeration area and enumerator
code for fixed effects.
To answer the second research question which is to identify if CAPI led to
a reduction in errors and if enumerator characteristics matter, the following
model was employed:

= + × + × + × + є

379 | I S I W S C 2 0 1 9