Page 386 - Special Topic Session (STS) - Volume 4
P. 386
STS2319 Lakshman N. R. et al.
How much better is better? Quantifying the CAPI
advantage using Viet Nam’s Labor Force Survey
2
1
Lakshman Nagraj Rao , Dave Pipon , Jude David Roque
*3
1 Asian Development Bank, Manila Philippines
2 Asian Development Bank, Manila Philippines
3 Asian Development Bank, Manila Philippines
Abstract
Labor statistics published by government agencies rely on data from Labor
Force Surveys (LFS), which in most countries are conducted using the pencil-
and-paper interviewing (PAPI) technique. More recently, there has been a
concerted effort for countries to switch to computer-assisted personal
interviewing (CAPI), wherein a handheld device is used during the interview
process. CAPI not only eliminates the need to manually re-enter the data, but
also automates questionnaire navigation and flags inconsistent responses on
the fly. While these features may lead to improvements in data quality and
timeliness, it is unclear to what extent, and whether these improvements affect
estimates during data analysis.
This paper presents results from a randomized experiment, designed
specifically to compare CAPI and PAPI using data from July 2017 - September
2017 for Ho Chi Minh in Viet Nam. Within each of a total of 180 sample
enumeration areas, 15 households were randomly selected and interviewed
using PAPI, while another 15 households were randomly selected and
interviewed using CAPI. This design allows for a detailed comparison of errors,
interview times, and costs between the two methods. In addition, we test the
hypothesis whether these errors are non-random, which may lead to
differences in estimates for basic labor force statistics between the two groups.
Keywords
Computer-assisted personal interviewing; data quality; randomized
experiment; survey; labor statistics
1. Introduction
Datasets matter in statistical analyses and tend to be scrutinized and
dissected down to minute details to generate meaningful results. Yet, not
much attention is given towards how such datasets are brought about in the
first place. The underlying data collection process along with the tools used,
associated errors, and implications for analysis are seldom considered. This is
important to consider because most data collection processes in developing
countries are still reliant upon traditional pencil/pen-and-paper interviewing
375 | I S I W S C 2 0 1 9