Page 182 - Contributed Paper Session (CPS) - Volume 3

CPS1985 Markus Z.
to get high quality official statistics will be done in the statistical offices in the
future in some cases semi-finished statistical products coming from private
date production will used. The Generic Statistical Business Process Model
(GSBPM) (https://statswiki.unece.org/display/GSBPM/GSBPM+v5.0) has to be
rethought in the context of IoT. The experience has shown that generally NSIs
aren’t able to produce real-time statistics based on sensor data alone
currently. A lot of investment is necessary for the IT infrastructure and
regarding the skills of the staff in the NSIs for this task. But the discussion has
only started whether this is efficient and possible.
The IT infrastructures in NSIs have to be further developed, but the
question is in which format. In some cases the necessary IT infrastructure is
established by the government, but in other official institutions. In some cases
buying data services could be cheaper compared to a situation in which NSIs
do the whole production by themselves. As discussed above, it is not clear
whether NSIs are able to hire data scientist at the labour market with the right
skills and in the quantity who is needed. The competition for getting these
skills is intensive and therefore this production factor is expensive, maybe too
expensive for NSIs. The next challenge is the access to the often privately held
data of the IoT. Current experience has shown that it is often more realistic to
buy the whole product as service; that means the product coming from the
combination of adequate IT infrastructure, skills and data. Often, private data
producers do not have the interest to sell only the detailed data. Generally,
buying this kind of semi-finished statistical products could be a good solution
for NSIs, because self-production is also expensive. The questions are, where
the break-even is, which quality the data have and how permanent the data
access is.
Future official data production probably will be done as work-sharing for
some products. Therefore, the rules for integration private data products into
official statistics are to define. Another important issue is the sharing of
responsibilities between the private and the official data production. First of
all, the statistical conception, including economic and sampling issues, is
needed for statistics production. This will still be the task of the statistician. But
statistical offices have to further develop their staff’s skills for doing these parts
of the statistical production. Probably it will not be the task of NSIs to do the
data engineering production steps, like developing algorithms to detect small
objects based on machine learning as example, but it will be necessary that
NSIs are able to understand the methods that are used. Besides, the
conception and the collection of the data and the publication of the results in
an interactive and intuitive format will be the third process step. The aim
should be to get a system of automated processes to steer the whole
production process free of media disruption and based on a further developed
GSBPM.

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