This study proposed a multivariate weighting system, which allows for the computation of both base and current year weights from price datasets conveniently. The proposed multivariate weighting system adopts a factor analysis methodology aimed at mitigating the widely reported weight and formula biases associated with CPI formulation, with an advantage of retaining variability in each price variable in the final index. The precisions of the expenditure-based weight (current weighting system) and that of the multivariate weighting system were assessed. It was found that the proposed weighting system performed better in terms of the proportion of indices falling within the ideal range. This result was consistent for 1000 bootstrap samples generated from 10 different multivariate data samples simulated for this purpose. The paper recommends the multivariate weighting system, because it is easy to implement and cost effective, as an alternative for the expenditure-based weighting system noted in literature for biases.
Keywords: index numbers; consumer price index; multivariate statistics; classical factor analysis
