Assessing Subjective Well-being in Wide Populations. A Posetic Approach. . .
247
phenomena, we must focus our attention on interrelations, rather than on causal
relationships among variables.
In this work we apply our analysis to data deriving from the ad-hoc module on
SWB of the European Union Survey on Income and Living Conditions (hereinafter:
Eu-SILC), adopted in the 2013 edition. These data are analysed by Eurostat in two
different reports. The first one (Eurostat 2015) examines data from this module,
aggregated at country level. It studies life satisfaction (hereinafter: LS) according
to some socio-economic dimensions (e.g. the working condition), and the meaning
of life (hereinafter: MoL) in relation to sex, age class and LS. The values of the
meaning of life are higher than those of LS, and the two variables (considering
data aggregated at the national level) seem to have an almost linear relationship,
with the Pearson 5 correlation coefficient (Eurostat 2013) of 0.56. Maybe, the
homogeneity in the structure of the two questions and the relative answers, as well as
the recoding of the two items into the same three classes, 6 could have influenced the
homogeneity of the distribution of the aggregate values. Then it appears necessary
to carry out a micro-data processing to better understand the relationships between
the two variables. As concerning the emotional state (hereinafter: ES), the module
adopted the MHI battery (from the SF-36 questionnaire) with questions relating
to five affects 7 . The questions have a very different structure compared to those
on LS and MoL, because they concern the frequency of each affect in a limited
period of time (four weeks). Furthermore, the five modalities of response 8 are
clearly ordinal. However, the Eurostat report analyzes only happiness, in relation
to age group, family structure, work condition, LS and MoL. The second Eurostat
report (Eurostat 2016) compares three multivariate regression models that consider
overall LS as a dependent variable. The third model includes mental well-being
as independent variable, considering the mental status 9 calculated on the MHI
battery – as a factor influencing SWB, and not as one of its components. None of
the two Eurostat reports considers the conjoint contribution of the three dimensions
in expressing the SWB level, although this is one of the recommendations of the
OECD guidelines. But above all, all the three variables expressing the dimensions
of SWB are analysed using statistical methods typical of cardinal variables, thus not
taking into account their ordinal nature.
In this paper, we present an application to the synthesis of SWB indicators taking
into account their ordinal nature. In particular, we propose a two-step synthesis
process: first, we address the synthesis of the indicators used to measure the ES
5 The choice of Pearson’s correlation coefficient reveals the questionable assumption that variables
are quantitative.
6 The two items are both revealed on a scale from 0 to 10. Recoding defined 3 modalities: low if
the level is less than six, median if the level is between 6 and 8 and high for levels 9 and 10.
7 Q: How much of the time, during the past four weeks have you been/felt: Very nervous; Down (in
the dump); Calm and peaceful; Downhearted and depressed; Happy.
8 A: All of the time; Most of the time; Some of the time; A little of the time; None of the time.
9 Mental status is calculated as the average score of the answers to the five questions on emotional
states, reporting the value on a scale from 0 to 100.
247
phenomena, we must focus our attention on interrelations, rather than on causal
relationships among variables.
In this work we apply our analysis to data deriving from the ad-hoc module on
SWB of the European Union Survey on Income and Living Conditions (hereinafter:
Eu-SILC), adopted in the 2013 edition. These data are analysed by Eurostat in two
different reports. The first one (Eurostat 2015) examines data from this module,
aggregated at country level. It studies life satisfaction (hereinafter: LS) according
to some socio-economic dimensions (e.g. the working condition), and the meaning
of life (hereinafter: MoL) in relation to sex, age class and LS. The values of the
meaning of life are higher than those of LS, and the two variables (considering
data aggregated at the national level) seem to have an almost linear relationship,
with the Pearson 5 correlation coefficient (Eurostat 2013) of 0.56. Maybe, the
homogeneity in the structure of the two questions and the relative answers, as well as
the recoding of the two items into the same three classes, 6 could have influenced the
homogeneity of the distribution of the aggregate values. Then it appears necessary
to carry out a micro-data processing to better understand the relationships between
the two variables. As concerning the emotional state (hereinafter: ES), the module
adopted the MHI battery (from the SF-36 questionnaire) with questions relating
to five affects 7 . The questions have a very different structure compared to those
on LS and MoL, because they concern the frequency of each affect in a limited
period of time (four weeks). Furthermore, the five modalities of response 8 are
clearly ordinal. However, the Eurostat report analyzes only happiness, in relation
to age group, family structure, work condition, LS and MoL. The second Eurostat
report (Eurostat 2016) compares three multivariate regression models that consider
overall LS as a dependent variable. The third model includes mental well-being
as independent variable, considering the mental status 9 calculated on the MHI
battery – as a factor influencing SWB, and not as one of its components. None of
the two Eurostat reports considers the conjoint contribution of the three dimensions
in expressing the SWB level, although this is one of the recommendations of the
OECD guidelines. But above all, all the three variables expressing the dimensions
of SWB are analysed using statistical methods typical of cardinal variables, thus not
taking into account their ordinal nature.
In this paper, we present an application to the synthesis of SWB indicators taking
into account their ordinal nature. In particular, we propose a two-step synthesis
process: first, we address the synthesis of the indicators used to measure the ES
5 The choice of Pearson’s correlation coefficient reveals the questionable assumption that variables
are quantitative.
6 The two items are both revealed on a scale from 0 to 10. Recoding defined 3 modalities: low if
the level is less than six, median if the level is between 6 and 8 and high for levels 9 and 10.
7 Q: How much of the time, during the past four weeks have you been/felt: Very nervous; Down (in
the dump); Calm and peaceful; Downhearted and depressed; Happy.
8 A: All of the time; Most of the time; Some of the time; A little of the time; None of the time.
9 Mental status is calculated as the average score of the answers to the five questions on emotional
states, reporting the value on a scale from 0 to 100.