Using Census data to derive a new area-based measure of deprivation - Appendix

Appendix: A further look into the application of SEISA

We argue that a key advantage of our measure is that it can be applied in a UK-wide analysis and for comparing statistics across the four nations. In this section, we summarise additional sensitivity checks we carried out regarding these features of SEISA. Specifically, we illustrate that where there are differences by nation in the way Census data is collected/disseminated, this does not seem to be resulting in the patterns we observe in the distribution of output/small areas. Furthermore, we note the consistency in the methodology used in generating SEISA across all four nations.

The ONS (2015) report highlights that key statistics on qualifications are broadly comparable, while they are highly comparable in the case of occupation. In the case of qualifications, although not discussed in the ONS (2015) publication, we do note that HNC/HND qualifications are placed in the ‘Level 3’ category within Scotland, but ‘Level 4’ in all other nations. To the best of our knowledge, the way in which HNC/HND qualifications have been grouped was determined based on user need from Census outputs.

Below, we provide a breakdown of how output/small areas are distributed across quintiles in each nation. We can see that, in the case of Scotland, 29% of output areas emerge in quintile 1 of the UK-wide measure. When utilising SEISA for a UK-wide analysis (i.e. a study which includes individuals from all nations), one may question whether the bottom quintile picks up a greater proportion of Scottish output areas due to the way qualifications have been banded across the various levels in Census outputs. However, to illustrate that this is not driven by the difference in the manner in which HNC/HND qualifications are grouped, we also provide a table illustrating how output/small areas are dispersed across the quintiles when based on the occupation variable only (which is highly comparable).

Evaluating the two tables (A1 and A2) demonstrates that the findings are very similar. As stated in the main paper, the very high linear correlation between qualifications and occupation means that there will not be a great deal of difference between a measure of deprivation based on both compared to one formed using one of the two variables only.

Table A1: The distribution of output/small areas across quintiles by nation for SEISA

Table A2: The distribution of output/small areas across quintiles by nation based on the Census occupation variable only

One may also hypothesise that the higher proportion of Scottish output areas that are observed in the bottom quintile could be due to Scottish output areas being smaller than those of other countries. To assess this, we carried out an additional analysis whereby we formulated our occupation variable (i.e. the most comparable Census field) using data zones in Scotland, but output/small areas in all other nations. In this scenario, it is Scotland that has the areas with the highest population sizes. The distribution by nation for the occupation field is demonstrated below. Even in this instance, a quarter of Scottish data zones fall within the bottom quintile. Consequently, it does not seem to be the case that output area size is the reason behind why a higher proportion of Scottish areas emerge in the bottom quintile.

Table A3: The distribution of output/small areas (data zones in Scotland) across quintiles by nation based on the Census occupation variable only

Please note that given this occupation variable was created using a different geographic domain in Scotland compared to the other nations, we have not supplied this field as part of the ‘Research dataset’ available through our interactive map webpages. Users who wish to recreate Table A3 could do so using the links/data provided in the ‘Research dataset’ for output/small areas in  England, Wales and Northern Ireland. This would need to be supplemented with occupation information for Scottish data zones (available at https://www.scotlandscensus.gov.uk/documents/2011-census-table-data-sns-data-zone-2011/).

Finally, for a measure of deprivation to be suitable for comparing statistics across nations, a key criteria that it should meet is that it is created using the same methodology across all countries. This is the case with SEISA. Indeed, as at a UK-wide level, whether one uses both qualifications and occupation to create the measure in a particular country or just the occupation variable is not likely to make any material difference to the analysis/results, given the strong linear correlation between the two fields at nation-level too. The highest correlations are observed in Northern Ireland and Wales (0.95 and 0.94, respectively), while the figures were 0.92 in England and 0.89 in Scotland.

Back: Summary