Page 60 - IJPS-2-2

Latent class models for cross-national comparisons: the association between individual and national-level fertility and partnership characteristics

multilevel modelling techniques (random and fixed effects) will often be severely limited
where countries are a higher unit, either due to the characteristics of the methods themselves or due
to the sampling structure at the country level. The implementation of latent classes at the country
level is an attempt to overcome the limitation of both of these methods. The additional advantage of
the latent class based approach is that in forming clusters based on empirical data, it provides the
opportunity to validate a priori theoretical clusters or typologies (e.g., Esping-Andersen, 1990; Esp-
ing-Andersen, 1999; Blossfeld, 2006).
That said there are a number of limitations to this analysis and a number of areas that could be
expanded. Firstly, the selection of the most recent variable information for policy indicators prec-
ludes causal inferences. The use of policy data from a point in time either contemporary or following
the life course processes under study introduces an ambiguity into the causal direction of the va-
riables under consideration: policies may influence life course behaviour; both the life course
processes and the policies pertaining to them may be endogenous with respect to broader cultural
trends within the country of study (McDonald, 2000), or policies may be derived to reflect social
pressure within the country (Perelli-Harris and Sánchez-Gassen, 2012). While it might at first seem
plausible to overcome this problem relatively simply by selecting historic indicators, the causal rela-
tionship is likely to be more complicated since policy will reflect underlying behavioural trends to
a certain extent (Perelli-Harris and Sánchez Gassen, 2012). However, similar to Neels et al. (2013),
it is not the aim of this paper to make causal statements about the particular economic variables in-
cluded; rather the indicators are taken as manifest representations of the generosity of the welfare
system in a country. It may be possible for researchers to apply this method to construct more robust
indicators and thus make causal statements, but this is beyond the scope of the current analysis. Se-
condly, the small number of countries can be considered a limitation in terms of power to ex-
tract classes of behaviour, and is somewhat less detailed than many conceptual typologies (Esp-
ing-Andersen, 1990; Esping-Andersen, 1999). That said, this ‘limitation’ is somewhat deliberate, as
much of the literature relies on a small country level sample. The fact that this article can produce
some meaningful interpretation despite this is still an advantage. Finally, the restriction of partner-
ship and fertility pattern to depend solely on the class allocation of the country is likely to be overly
simplified. Indeed, similar techniques have found multiple classes at an individual (sub-national)
level (Dariotis, Pleck, Astone, et al., 2011; Perelli-Harris and Lyons-Amos, 2015; Perelli-Harris and
Lyons-Amos, 2016). The production of sub-national level classes and their cross classification with
national level policy classes is a natural extension of this paper.
The advantages of this approach are demonstrated by evaluating the association between fertility
and partnership behaviour and country clusters represented by a set of policy relevant indicators.
This analysis extracted three distinct country level clusters in the European settings. Firstly, Eastern
European countries were distinct exhibiting relatively low welfare support in both absolute and rela-
tive terms, but with a fair degree of child support. There was no protection for cohabitation as
a childbearing union. Unions tended to be marital in these settings, with an early incidence of fertili-
ty behaviour. The uniqueness of the Eastern fertility pattern and extraction as a discrete cluster dis-
tinct from Western fertility regimes has been remarked upon in demographic literature and links to
theoretical understanding of fertility patterns (Blossfeld and Drobnic, 2001).
Western Europe was characterised by two classes: lesser and greater support. The lesser
port class (The Netherlands and Spain) comprised countries in Conservative and Southern welfare
state typologies; both characterised by welfare systems designed to either support the family institu-
tion or rely on familial support (Blossfeld, 2006). The uniqueness of the Netherlands as a hybrid
welfare state model has been noted elsewhere (Kammer, Niehues, and Peichl, 2012) and seems to be
replicated by this analysis. This class had relatively low financial support for childbearing and wel-
fare in general, but relatively high levels of female labour market engagement. There was some li-
mited legal protection for cohabiting unions. Again, in terms of partnership behaviour, unions are

54 International Journal of Population Studies | 2016, Volume 2, Issue 2