Nonparametric analysis of shape and dynamics of probability distributions has recently been used to identify clustering and polarization phenomena in regional economic development (Quah, 1997 a, b). The technique used to investigate which variables are relevant in determining such polarization phenomena is known as conditioning. So far this kind of approach has been applied only in a univariate context. In this paper, we propose a more general framework, referring to the literature on nonparametric local regression techniques. Following this alternative approach we are able to study the joint effect of two or more variables on the shape of the distribution of growth. An empirical illustration is given for the distribution of employment rate in 784 Italian Local Labour Markets.