The effects of measurement error in cross cultural research

Abstract

"In survey research many decisions are made in order to design an instrument for data collection. These choices have to do with the formulation of the question, the response categories, the instruction, the sample, the mode of data collection, etc. Each of these choices can lead to different errors (Sudman and Bradburn, 1974; Belson, 1981; Schuman and Presser, 1981; Dijkstra and Van der Zouwen, 1982; Andrews, 1984; Molenaar, 1986; Billiet et al., 1986; Groves, 1989; Alwin and Krosnick, 1991, and Scherpenzeel and Saris, 1997) and consequently to incomparability of results with respect to estimates of correlations and effect parameters across studies and also across countries. It is common knowledge that cross-cultural comparison can only be made if the measurement procedures are completely the same. In this study, we want to argue that this requirement is not enough. We will show that the results can also differ if the same procedures have been used because of differences in measurement errors in the different countries. We therefore propose a procedure to correct for measurement error, in order to make comparisons across countries with respect to correlations and regression coefficients. To correct for measurement error, we have chosen an approach that can be used by every researcher involved in social science research. This in particular is why we advocate this approach, even though, from a methodological point of view, more suitable approaches are available. We avoid using these methods because one purpose of this project is that we want to demonstrate a procedure for the correction of measurement error which can be used in any study, once prior methodological research is done. We begin with a discussion of the problems connected with measurement error in comparative survey research and then we describe the solution we have chosen for these problems. All examples given are based on the satisfaction studies done in the context of a methodological, comparative research project involving 13 language areas." (author's abstract)