246
L. S. Alaimo and P. Conigliaro
of non-elementary concepts (OECD 2013). Moreover, the correlations between the
three main dimensions are not so strong (Clark and Senik 2011; Huppert and So
2013) as to allow ignoring the contribution of one of them in the evaluation of
SWB. This confirms the need to reveal all three main dimensions of SWB and to
define interpretative tools that can safeguard the multidimensionality of the concept
(OECD 2013).
Comparing different groups of respondents requires synthesizing the information
gathered on very large databases. By doing this, surveys become accessible and
usable to a wider public, supporting democratic participation and political decisions.
However, the need to reduce the complexity can lead to the risk of excessive
simplification of the concepts and to distortive interpretations of the information
collected. In particular, the identification of the measurement model and the choice
of the synthesis method are critically important steps, which have an impact on the
results. An explicit or implicit conceptual model always guides the interpretation
of the relationship between indicators. According to Maggino (2017), the main
distinction is between reflective and formative models. The reflective model refers
to a latent variable that exists irrespective of the units of measurement and the
units of analysis. The indicators must intercept it, interpret it, and measure it. This
means that the indicators are interchangeable and that the internal consistency is of
fundamental importance: if two indicators are uncorrelated, they do not measure
the same concept. The psychometric tools, usually adopted to measure SWB in
statistical surveys, use a reflective model. The indicators’ effectiveness consists
in their capacity to approximate the measure of a characteristic (mental health,
intelligence, ability). In the study of social phenomena, the most common approach
is, instead, the formative (or constructivist) one. According to this model, the
indicators do not depend on the latent variable, but they determine its nature and
characteristics. The formative is a bottom-up explanatory approach. The internal
consistency is of minimal importance: two uncorrelated indicators can both be
useful and even indispensable for the knowledge of multidimensional phenomena.
Therefore, omitting an indicator means omitting part of the construct. In this paper,
we deal with a formative measurement model 4 .
The correct definition of the conceptual model allows to correctly interpret
the relationships between the indicators and to correctly identify the procedures
for synthesizing them (Maggino 2017). Thus, it is important to use the correct
method of synthesis, considering the nature of the data. As previously written,
many studies process ordinal information as a continuous quantitative variable,
using then synthesis methods traditionally adopted for this type of variable (e.g.
the arithmetic mean). We consider these methods unsuitable to synthesize ordinal
data and misleading for conclusions. Furthermore, in the analysis of complex
4 It should be made clear that the choice of the measurement model does not depend on a free
choice of the researcher, but exclusively on the nature of the latent variable measured (Alaimo and
Maggino 2020).
L. S. Alaimo and P. Conigliaro
of non-elementary concepts (OECD 2013). Moreover, the correlations between the
three main dimensions are not so strong (Clark and Senik 2011; Huppert and So
2013) as to allow ignoring the contribution of one of them in the evaluation of
SWB. This confirms the need to reveal all three main dimensions of SWB and to
define interpretative tools that can safeguard the multidimensionality of the concept
(OECD 2013).
Comparing different groups of respondents requires synthesizing the information
gathered on very large databases. By doing this, surveys become accessible and
usable to a wider public, supporting democratic participation and political decisions.
However, the need to reduce the complexity can lead to the risk of excessive
simplification of the concepts and to distortive interpretations of the information
collected. In particular, the identification of the measurement model and the choice
of the synthesis method are critically important steps, which have an impact on the
results. An explicit or implicit conceptual model always guides the interpretation
of the relationship between indicators. According to Maggino (2017), the main
distinction is between reflective and formative models. The reflective model refers
to a latent variable that exists irrespective of the units of measurement and the
units of analysis. The indicators must intercept it, interpret it, and measure it. This
means that the indicators are interchangeable and that the internal consistency is of
fundamental importance: if two indicators are uncorrelated, they do not measure
the same concept. The psychometric tools, usually adopted to measure SWB in
statistical surveys, use a reflective model. The indicators’ effectiveness consists
in their capacity to approximate the measure of a characteristic (mental health,
intelligence, ability). In the study of social phenomena, the most common approach
is, instead, the formative (or constructivist) one. According to this model, the
indicators do not depend on the latent variable, but they determine its nature and
characteristics. The formative is a bottom-up explanatory approach. The internal
consistency is of minimal importance: two uncorrelated indicators can both be
useful and even indispensable for the knowledge of multidimensional phenomena.
Therefore, omitting an indicator means omitting part of the construct. In this paper,
we deal with a formative measurement model 4 .
The correct definition of the conceptual model allows to correctly interpret
the relationships between the indicators and to correctly identify the procedures
for synthesizing them (Maggino 2017). Thus, it is important to use the correct
method of synthesis, considering the nature of the data. As previously written,
many studies process ordinal information as a continuous quantitative variable,
using then synthesis methods traditionally adopted for this type of variable (e.g.
the arithmetic mean). We consider these methods unsuitable to synthesize ordinal
data and misleading for conclusions. Furthermore, in the analysis of complex
4 It should be made clear that the choice of the measurement model does not depend on a free
choice of the researcher, but exclusively on the nature of the latent variable measured (Alaimo and
Maggino 2020).