Page 283 - Special Topic Session (STS) - Volume 2
P. 283
STS493 Sofie d.B. et al.
Sensor data at the heart of innovation in official
statistics
1
Sofie de Broe , Ger Snijkers , Barry Schouten 12
1
1 Statistics Netherlands, Statistics Netherlands and Utrecht University
2
Abstract
This paper deals with two approaches of using sensor data in official statistics.
Both approaches lead to hybrid data collection in which sensor data is
combined with additional survey data. The first approach invites sample units
to collect sensor data through mobile devices, wearables and/or IoT type
sensors. The second approach assumes that sensor data have already been
collected by sampling units and these units are asked for consent to use these
data for a specified time period. The two approaches can be applied to both
business statistics and social statistics. However, the type of sensors that are
employed and the resulting sensor data are typically different for the two
target populations. In the paper, we introduce criteria to determine the utility
of sensor data for official statistics. We do so from the perspective of output,
from the perspective of the sensors and from the perspective of the target
populations. We illustrate the criteria for two realistic case studies; one for
social statistics and one for business statistics.
Keywords
Sensor data; Mobile device; Wearable; Survey; Measurement equivalence
1. Introduction
Sensor data have become omnipresent in business processes and in daily
life. Mobile devices have become standard tools for communication, daily
archiving and personal administration within the time span of just a decade
and have a high population coverage worldwide. Simultaneously, authorities
and businesses have implemented all kinds of sensors for monitoring and
improving of events and activities. Because of their high population coverage
and daily life use, mobile devices, wearables and IoT sensors have become
tools that may supplement surveys with automated data from sensors. Some
of these sensor data may also replace survey data or may even have the
potential to introduce revised concepts and views on statistics.
Naturally, the potential to record or link sensor data is not sufficient reason
to also do so. Criteria are needed that identify combinations of survey topics
1 The authors like to thank Ole Mussmann for his input on mobile device and wearable sensors
and Tim Punt for his help in applying the various criteria to the agriculture case study.
272 | I S I W S C 2 0 1 9