Detecting heteroscedasticity using a non-parametric regression technique

Most empirical research on detecting heteroscedasticity relies on parametric techniques. It is, however, well established that parametric estimation techniques have the added disadvantage of introducing biases from parameters fed into the modelling process. In this paper, a diagnostic test for detecting heteroscedasticity is explored using a non-parametric approach. The small and large sample statistics of the hypothesis of homogeneous variances are also studied using stock FTSE 100 index data. The test statistics rejected the null hypothesis of homogeneous variances.