ACSPRI Conferences, RC33 Eighth International Conference on Social Science Methodology

Font Size:  Small  Medium  Large

Multiple Correspondence Analysis: An Alternative for Preserving Complexity in Value Research

Zoltan Lakatos

Building: Law Building
Room: Breakout 2 - Law Building, Room 026
Date: 2012-07-10 11:00 AM – 12:30 PM
Last modified: 2012-04-03


Correspondence analysis is rarely applied in research into cultural values. Given that early applications of CA by social scientists (most significantly by Bourdieu, 1979) were made in the field of cultural sociology, this lack of interest is surprising. The one-sided focus on techniques like principal component and factor analysis may be due to researchers’ preference for efficient methods to deal with cross-country survey data. Multivariate analyses are often performed on national averages?an approach that its proponents justify on grounds of getting rid of measurement error. This apparent efficiency comes at a heavy price: methods not suited to categorical data, often on top of the ecological fallacy may lead to wrong conclusions about not only cross-country differences regarding values but also the very content of these constructs.
Because most questions in value research concern individual behavior and input variables are typically categorical rather than ordinal, multiple correspondence analysis provides a sensible alternative. The study presented in this paper performs MCA on data used by Inglehart (1995, 2000) to construct his typology of cultural values and also an alternative set of variables (both from the World Values Survey) in an effort to address the major issues with Inglehart’s method, widely discussed in the literature. It follows on work by authors critical of Inglehart’s methods, most importantly Flanagan (1982), Haller (2002), and Majima and Savage (2007) who first proposed MCA as an alternative to overcome their inherent reductionism. It applies correspondence analysis with orthogonal rotation using two recently developed software packages: CAR for MATLAB by Lorenzo-Seva, van de Velden and Kiers (2007) and PCAmix for R by Chavent, Kuentz, Liquet and Saracco (2011).
MCA unravels the complexity of cultural values largely unnoticed by umbrella constructs resulting from factor analysis performed at the aggregate level. Among others, it sorts out psychological variables that are not related to values but which, together with an indicator of postmaterialism account for much of the “survival vs. self-expression” dimension in Inglehart’s analysis. Orthogonal rotation separates out authoritarianism from religiosity?two dimensions that Inglehart’s “traditional vs. secular” typology presents as if they were part of the same latent construct and even unrotated MCA solutions identify with one single axis. It also isolates achievement orientation and materialism as further dimensions. I argue that failure to recognize how individuals are positioned in such a complex field has led to a number of misconceptions about the direction of value change.