Page 290 - Invited Paper Session (IPS) - Volume 2
P. 290
IPS246 Tiziana Laureti et al.
paramount . Indeed, it is crucial to take a measure of uncertainty into
1
account when monitoring poverty in order to avoid misleading regional
analyses and the consequent policy implications and outcomes.
Even if the importance of providing measures of uncertainty of poverty
indicators has been widely recognized, NSOs are reluctant to communicate
uncertainty measures of reported estimates in news releases while technical
publications acknowledge that these statistics are subject to error by
providing information on the methods used for estimation.
As EU-SILC is a sample-based survey, both sampling and non-sampling
errors can seriously affect the accuracy of all estimates derived from this
survey (Verma et al, 2010). However, the computation of standard errors
for estimates of poverty indicators based on EU‑ SILC is a challenging task
due to the complex sample designs employed by several EU countries
which involves stratification, geographical clustering, unequal probabilities
of selection for the sample units and post-survey weighting adjustments
(re-weighting for unit nonresponse and calibration to external data sources
for non-response adjustment of the initial design weights). Additional, EU-
SILC has an important panel component with a 4-year rotational panel
design in the majority of countries. Moreover, full documentation of the
sample design and accurate sample design variables in the EU-SILC dataset
are usually lacking, thus hampered research studies focused on exploring
sampling error measures.
Several methods for estimating the variance of the poverty indicators
have been discussed in the literature (Berger et at, 2017) These methods
that can be classified into two approaches: ‘direct’ methods, which rely on
analytic variance formulas through linearization (Alper and Berger, 2015),
and ‘resampling’ methods, such as Jackknife repeated replication or
boostrap, which consist of re-sampling a high number of ‘replications’ from
the original sample in order to empirically derive a sampling distributions
(Davidson and Flachaire, 2007; Verma et al, 2010). Contrastingly, standard
error and confidence intervals of regional poverty indicator has been
studied only in a limited number of papers (Verma et al 2017) although
they are needed for regional (subnational) estimates for assessing and
monitoring regional policies. In this context, the main difficulty arises from
the smallness of regional samples in national surveys. Various methods
could be used to overcome this issue as illustrated by Verma et al 2017.
However, it is worth noting that whichever approach is used, in order to
obtain reliable standard error estimates it is essential that the sample
1 By referring on the content of intermediate and final EU‑ SILC Quality reports, Commission
Regulation (EC) No 28/2004 of 5 January 2004 requires that countries should provide estimates
of standard error along with the EU‑ SILC main target indicators.
277 | I S I W S C 2 0 1 9