Background

The Survey of Income and Living Conditions (SILC) is weighted by estimates of population and the number of households at NUTS3 regional level. Because of EU regulation changes, the NUTS3 region classifications used for weighting have changed with effect from January 1^st, 2018. Following the publication of the Census of Population 2016, annual population estimates for Ireland from 2012 onwards were revised.  With the publication of the SILC 2017, the existing SILC series from 2012 to 2016 has also been reweighted to account for the new NUTS3 groupings and new population estimates to facilitate the availability of a time series of data for users of the SILC data.

Revisions to composition of NUTS Regions

The NUTS (regional classification) boundaries were amended on 21^st November 2016 under Commission Regulation (EU) No 2016/2066 and took effect from 1^st January 2018.  Please see http://ec.europa.eu/eurostat/web/nuts/history for further details.

As a result, new NUTS groupings have been introduced for Ireland. At NUTS3 level, County Louth has moved from the Border to the Mid-East region and what was formerly South Tipperary has moved from the South-East to the Mid-West region. At NUTS2 level, the change sees the number of regions increase from two to three. Please see Appendices I and II for the new and old classifications.

The Central Statistics Office (CSO) weights the SILC using NUTS3 regional groups and is required under EU law to publish the SILC for 2017 using these new NUTS3 region classifications.  Estimates of population and the numbers of households according to the new NUTS3 regional groupings are only available back to 2012 for Ireland.  The CSO has revised the SILC results back to 2012 to provide the maximum possible regional time series to our users.  This reweighted data from 2012 to 2016 inclusive is published with the SILC 2017 results and users should note that there is a break in the regional data series from 2012 as the results for the period 2004 to 2011 are published using the old NUTS groupings.

Census of Population 2016

As the SILC is a sample survey, independent estimates of population and numbers of households are required each year to provide a weighting basis for the statistics produced from the SILC.  The process of deriving these population and household totals involves using the population counts from the most recent Census of Population as a base, and updating these each quarter using information on births, deaths and migration for that quarter.  The SILC then takes the average of the four quarterly estimates of population and numbers of households from each calendar year to provide annual estimates of population and numbers of households which are then used for weighting the SILC results. 

Once a new Census of Population has been completed, the quarterly population estimates back to the previous Census of Population are revised.  The Census of Population 2016 was conducted on April 24^th, 2016 which means that it fell in Quarter 2 (Q2) 2016.  The difference between the new population estimates for Q2 2016 following the release of the Census 2016 results and the original population estimates for Q2 2016 is in the order of 65,000. 

Prior to the publication of Census of Population 2016 results, the previous benchmark for population estimates was Census of Population 2011 (Q2 2011) and the population estimates for each quarter since then have now been revised.  There was a cumulative element to the revisions with relatively small differences in the earlier periods with the difference increasing over time, i.e. the population estimates closer to Q2 2011 needed to be revised to a lesser extent than later quarters and years.

The revised quarterly (post Census of Population 2016) population estimates for Q3 2011 onwards became available in late 2017 following the publication of the Population and Migration Estimates April 2017 by the CSO in September 2017. Following their release, the Labour Force Survey (LFS) results for all quarters from Q3 2011 onwards were reweighted according to these new population estimates when the Q3 2017 LFS release was published in January 2018.  All benchmark files containing household weights from the LFS to be used for weighting the SILC for 2017 and onwards will be based on these post Census of Population 2016 population estimates.

As mentioned above, given the critical importance of statistics on income and living conditions, the CSO took the decision to provide the maximum possible regional time series for the SILC using the new NUTS3 regions, i.e. back to 2012.  The CSO is also taking this opportunity to reweight the SILC from 2012 through to 2016 using the updated population estimates. Therefore, the results published for the SILC 2017 incorporate the new estimates of households and population using the new NUTS3 regions for each year since 2012 into the weighting methodology.

Increased Use of Administrative Data

The Income, Consumption and Wealth (ICW) Division within the CSO is currently preparing for a new European regulation; the Integrated European Social Statistics (IESS) regulation. The IESS regulation is a framework regulation to produce European statistics on persons and households and is expected to come into effect in 2021 or 2022.  As part of preparatory work for the IESS regulation, the ICW division is working on ways to reduce respondent burden and improve the quality of the data collected by the SILC.

As results for the SILC from 2012 through to 2016 are being revised due to the new NUTS3 regional classifications and the post Census 2016 population and household estimates, the ICW division in the CSO took the opportunity to increase the use of administrative data in the SILC process.  The process changes mostly relate to increased usage of administrative data for employee income variables and this has improved the quality of the data from 2012 onwards.  While applying the process changes to the periods in question a number of corrections were made to further improve the quality of the data.  The revised estimates for at risk of poverty, consistent poverty, enforced deprivation, Gini coefficient and quintile share ratio are not statistically significantly different from the pre-revision estimates.