22
F. Maggino et al.
strated empirically by testing the selected model of measurement. The specification
of the latter refers to the relationship between constructs and indicators. In literature,
we distinguish two different conceptual approaches: reflective and formative. 1
3. Basic indicators
In the majority of the cases the defined variable can be measured only indirectly
through observable elements which are called indicators of the reference variable.
Each basic indicator represents what can be actually measured in order to investigate
the corresponding variable. In other words, the indicator is what relates concepts to
reality.
We must specify that in all phases of hierarchical design is involved subjectivity.
Measurement is not an arbitrary process, but necessarily involves subjectivity.
There will always be the influence of the subject’s point of view: in the definition
of phenomena; in the definition of the hypotheses on reality; in the selection
of indicators; in the choice of statistical tools. Subjectivity represents one of
the dimensions inevitably involved in defining concepts, making measurement a
complex exercise.
The proper and accurate application of the hierarchical design allows defining
a complex structure in which each indicator measures and represents a distinct
component in the description of the phenomenon. Different types of indicators can
be present within a system, contributing to its complexity.
Within a system, indicators may show different characteristics related to (i)
the perspective through which the indicators are reporting the phenomenon to
be observed, (ii) the level of observation (e.g. micro/macro, internal/external),
(iii) the nature of the observed characteristics (e.g., objective/subjective, qualitative/quantitative), (iv) the level of dis/aggregation, (v) the communication context
in which the indicators are used, (vi) the interpretation attributed to the indicators in
statistical analyses, (vii) the criteria of their adoption, and (ix) their quality. 2
Indicators are also classified according to the type of data they contain. We can
have continuous (metrics), discrete (counts, we can perform operations of a linear
space can be performed, if meaningful, such as weightings, weighted sum as utility
function), ordinal (ratings/ranks, for which at least comparisons are possible) and
nominal data (descriptive, but restricted applicability, for instance stratifications).
One important question we have to address is how many indicators there should
be within a system. There is no single answer. Choosing too few of them carries
1 For the main characteristics of these two different approaches, see Maggino (2017a). The
literature about the difference between these two tipes of models is rich. As shown by Alaimo
and Maggino (2020), the state of the theory on formative models has been in intense discussion for
some years. Several authoritative scholars (for instance, Edwards 2011; Aguirre-Urreta et al. 2016)
have questioned the validity of this method and published appeals to no longer host its applications
in scientific journals. The debate seems to be far from being resolved. We would like to point out
that the choice between the two types of model does not depend directly on the researcher, but
exclusively on the nature and direction of relationships between constructs and measures.
2 For more information, see Maggino 2017a.
F. Maggino et al.
strated empirically by testing the selected model of measurement. The specification
of the latter refers to the relationship between constructs and indicators. In literature,
we distinguish two different conceptual approaches: reflective and formative. 1
3. Basic indicators
In the majority of the cases the defined variable can be measured only indirectly
through observable elements which are called indicators of the reference variable.
Each basic indicator represents what can be actually measured in order to investigate
the corresponding variable. In other words, the indicator is what relates concepts to
reality.
We must specify that in all phases of hierarchical design is involved subjectivity.
Measurement is not an arbitrary process, but necessarily involves subjectivity.
There will always be the influence of the subject’s point of view: in the definition
of phenomena; in the definition of the hypotheses on reality; in the selection
of indicators; in the choice of statistical tools. Subjectivity represents one of
the dimensions inevitably involved in defining concepts, making measurement a
complex exercise.
The proper and accurate application of the hierarchical design allows defining
a complex structure in which each indicator measures and represents a distinct
component in the description of the phenomenon. Different types of indicators can
be present within a system, contributing to its complexity.
Within a system, indicators may show different characteristics related to (i)
the perspective through which the indicators are reporting the phenomenon to
be observed, (ii) the level of observation (e.g. micro/macro, internal/external),
(iii) the nature of the observed characteristics (e.g., objective/subjective, qualitative/quantitative), (iv) the level of dis/aggregation, (v) the communication context
in which the indicators are used, (vi) the interpretation attributed to the indicators in
statistical analyses, (vii) the criteria of their adoption, and (ix) their quality. 2
Indicators are also classified according to the type of data they contain. We can
have continuous (metrics), discrete (counts, we can perform operations of a linear
space can be performed, if meaningful, such as weightings, weighted sum as utility
function), ordinal (ratings/ranks, for which at least comparisons are possible) and
nominal data (descriptive, but restricted applicability, for instance stratifications).
One important question we have to address is how many indicators there should
be within a system. There is no single answer. Choosing too few of them carries
1 For the main characteristics of these two different approaches, see Maggino (2017a). The
literature about the difference between these two tipes of models is rich. As shown by Alaimo
and Maggino (2020), the state of the theory on formative models has been in intense discussion for
some years. Several authoritative scholars (for instance, Edwards 2011; Aguirre-Urreta et al. 2016)
have questioned the validity of this method and published appeals to no longer host its applications
in scientific journals. The debate seems to be far from being resolved. We would like to point out
that the choice between the two types of model does not depend directly on the researcher, but
exclusively on the nature and direction of relationships between constructs and measures.
2 For more information, see Maggino 2017a.