Page 292 - Invited Paper Session (IPS) - Volume 2

IPS246 Tiziana Laureti et al.
computations can be performed for each region in the same way as at
the national level.
Since the poverty threshold chosen can influence the resulting
poverty rates when there is a high degree of disparity between the units
of analysis, as in the case of Italy, we estimated 95% confidence intervals
for AROP using two different procedures based on a national poverty
line estimated from the sample by means of: a) equivalised household
income and b) equivalised household income adjusted for price
differentials among regions using as proxy of the overall sub-national
Spatial Price Indexes (SPIs) the “Food products” SPIs calculated using
scanner data from modern retail chains. These SPIs are obtained within
an ISTAT research project for computing sub-national SPIs based on
scanner data and CPI data (Laureti and Polidoro, 2018). Even if SPIs for
food consumption aggregate represents only a part of the total
household consumption expenditure it may be interesting to analyse
what happens when price dimension is included into AROP standard
error measurement. Indeed, in this case an additional source of
uncertainty is introduced. Further research will be devoted to this issue
within the COMUNIKOS project.
Another critical aspect in measuring uncertainties in poverty
indicators is how to communicate them in a “comprehensive” way, in
terms of capturing fully the uncertainties, but also in a “understandable”
way so that different users and readers of these data correctly infer and
interpret the uncertainties communicated to them. Increasing attention
has been paid to this aspect in literature (Spiegelhalter et al., 2011; van
der Bles et al., 2019). COMUNIKOS aims also to carry out a detailed
investigation of the pros and cons of communicating uncertainties to
users of official statistics by considering appropriate tools for measuring
and disseminating data uncertainties.

3. Results
With the aim of providing an idea of the role visualization when
communicating uncertainty in poverty indicators, we explore various ways for
displaying uncertainty using bar charts (van der Laan, 2015). Figure 1a depicts
the point estimates for AROP without information on standard errors, as it is
usual practice, while Figure 1b shows only 95% confidence intervals for Italian
regions. Figure 2a and Figure 2b show AROP point estimates together with
95% confidence intervals for Italian regions. In these figures bar chart with
error bars are compared with bar chart with cross bar. While the first chart puts
a visual focus on the point estimate the cross-bar chart puts more emphasis
on the uncertainty measure. Crossbar chart seems to allow a clearer

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