Innovation as a latent variable: an alternative measurement approach

In searching for improvements in the measurement of innovation, in this paper we suggest the idea of conceiving innovation as a latent variable or concept, i.e., one that cannot be accurately defined or directly measured. We propose the use of multivariate statistical analysis (a combination of cluster and discriminant analysis) as a possible way to develop indirect measures of innovation that are more appropriate to the intrinsically imprecise nature of the concept to be measured. In this way, the problem of using an ambiguous and subjective ‘innovation definition’ is limited, reducing measurement errors. We obtain an ‘innovation intensity index,’ which is a weighted average of ten different predictors.