ABSTRACT
This chapter analyses a situation where some or all of the explanatory variables in the randomly varying coefficients model are collinear. Multicollinearity refers to a situation where perfect interrelationships exist among the explanatory variables. The several possible methods suggested for solving multicollinearity problems are: getting more data; making use of extraneous estimates; dropping some of the explanatory variables; using principal components; using ridge regression and using priori information and employing Bayesian estimation. A motivation for specifying the wage-price behaviour with randomly varying coefficients is that the short run Phillips curve is inherently unstable. The chapter discusses an illustration of the two stage random ridge (2SRR) estimation, and considers the estimation of a non-homogeneous Cobb-Douglas production function with purely random coefficients. The existence of a long run inflation-unemployment trade-off relationship suggests that the operational Phillips curve is an average relationship.
