THE QUANTIFICATION OF LATENT VARIABLES IN THE SOCIAL SCIENCES: REQUIREMENTS FOR SCIENTIFIC MEASUREMENT AND SHORTCOMINGS OF CURRENT PROCEDURES

摘要

In the social sciences, latent constructs play an important role. They appear as explanatory elements in structural theories, or they are of interest as the outcome of an intervention, for example a support or a preventive programme, a therapy, or a marketing activity. These constructs are typically considered to imply a quantitative latent variable that exists independently of measurement. As a matter of routine, measures of latent variables in the social sciences are treated in the same way as natural scientists handle and utilize their measures. However, in terms of what the concept of measurement is actually about, the social sciences have veered away from the rigorous concept adhered to in the natural sciences. An arbitrary definition of measurement and a multitude of procedures which are deemed appropriate for quantification have resulted in a speculative approach to measurement. Based on a return to the standard definition of measurement and a new conceptualisation of content and construct validity, the social sciences could advance their quantitative research substantially. The Rasch model for measurement plays an important role in this process.