Each higher order model for the three variables, namely globalization, consumption patterns, and culture is good fitted because the values of all of the fit indices are according to the defined limits (mentioned in Table 2).
Higher order models are applied because of the fact that first order latent variables of all three main latent variables were correlated.
They used data from national probability samples of citizens from the US and Canada to test a higher order model of environmental concern.
More importantly, we have found that: (a) a two-dimensional higher order structure tend to fit the data better than a structure with a single higher order factor; (b) the factors of a two-dimensional higher order structure corresponded to Preservation and Utilization dimensions of environmental attitudes; (c) discriminant validity for these two higher order dimensions was demonstrated by showing that self-reported ecological behaviour was predicted by Preservation, and not by Utilization, while attitudes toward economic liberalism were predicted by Utilization, and not by Preservation, and (d) this two-dimensional higher order model of environmental attitudes has also theoretical support (Milfont, 2007b).
However, there are also empirical and theoretical evidence supporting a single higher order model (cf.
Overall, the evidences gathered so far seems to indicate that the single higher order model is preferable on the grounds of the principle of parsimony.