Subjective Well-Being in Italian Regions

Abstract

This study presents a comparative analysis of subjective well-being in the Italian regions (NUTS2) by synthesizing several ordinal indicators. In recent decades, the use of subjective indicators, i.e. information collected at individual level, has become widespread in official statistics. This article refers, in particular, to those subjective indicators that detect perceptions, points of view and emotional states. The data source is the ad-hoc module on subjective well-being of the European Union Survey on Income and Living Conditions, one of the harmonised European official statistical surveys. In synthesizing the information, the study applies the Partially Ordered Set methodology. The purpose is to safeguard the multidimensionality of the phenomenon and the ordinal nature of the items. The study synthesizes the dimension of subjective well-being at micro level, i.e. at individual level, and reports the individual condition of subjective well-being at regional level. In doing so, it explores the informative potential of the ad-hoc module on subjective well-being, which seems not yet fully exploited.

Appendices

Appendix 1

See Table 2.

Table 2 Variables in the EU-SILC ad-hoc module 2013, characteristics of the questions and some dimensional information

Appendix 2

2.1 Effects on Synthetic Results, When Changing an Indicator or the Threshold

A relevant issue concerns the influence of researchers’ choices on the final results of the synthesis. This influence is mainly exercised through two main interventions: the choice of indicators and the definition of thresholds.

As we have seen the choice of indicators does not change the structure of the graph nor the position values, if the variables have the same number of modalities. In fact, in a set of four indicators each one articulated in three modalities, the possible combinations always being 81. In all the sets conformed in this way an identical profile will manifest identical values of average rank, identification, severity and wealth. This is what was meant when we talked about the indifference of the graph structure to the content of the variables.

Table 3 shows the results of a comparison between two different sets regarding the cognitive dimension. The first set (SAT-1) is the one presented in the article, the second (SAT-2) replaces the variable satisfaction for living environment with the variable satisfaction for accommodation. Both are expressed in three modalities and their frequency distribution in the total population registers a better condition for housing satisfaction. In the two sets the parameters reported in columns 2–7 are identical, so they are not repeated. What distinguishes the final values is the distribution of respondents by profile (column 8 and 9 in yellow in the table). This may cause a difference in the general levels of the poverty gap and wealth gap parameters. This difference is highlighted in Table 4.

Table 3 Effects of changing one indicator or changing threshold–Cognitive dimension

Table 4 Poverty gap and real gap in the three different options

Table 3 also shows the comparison of the results of the evaluation function if the threshold is changed. Changing the threshold means defining the deprived profiles differently. In this article the chosen threshold for the set SAT-1 is (a) [2-2-2-2]. What happens if we choose another threshold? In the example in Table 3, the new threshold is (b) [3-2-2-2]. Comparing columns 10, 11 and 12 (in light blue) with columns 2, 4 and 5, we can see the differences for the parameters of the identification functions, relative severity and relative wealth gap.

In Table 4 there is a comparison of the different levels of poverty gap and wealth gap, generated introducing these variation in the set of indicators or in the threshold.

Differences become more evident when we change the threshold. They are, however, less evident than those registered between the different regions. This leads us to suppose that the method is more sensitive to the respondents’ evaluations than to the researcher’s choices.