A Compound Gauss-Markov Random Field (CGMRF) modeling of Philippine unemployment data

Abstract

The presence of discontinuities in the January rounds of Philippine Unemployment Data from 1981–2006 is investigated by way of modeling these data as a noisy Compound Gauss-Markov Random Field (CGMRF). The likelihood and prior hyperparameters are respectively estimated with wavelet shrinkage and least squares. The posterior random field signal and its line process were obtained using the Gibbs Sampler and were optimized by Simulated Annealing. Results indicate that there has been an abrupt change in unemployment rates in the mid-1980s and in 2005–2006. The former may reflect the political and economic crisis that followed the Aquino assassination, while the latter changes reflect the new official definition of unemployment adopted in 2005.