Censored Latent Effect Autoregression Model on Canadian Unemployment Rate: A Non-linear Approach

Non-linear economic methods are often employed to describe the asymmetry of unemployment rates in the industrialized countries during the postwar period. Perhaps most common are Markov switching models. However, this project employs a different approach, a so called "censored latent effects autoregression" (CLEAR) model to describe the data. After specification tests, a parsimonious model is suggested to produce a good fit of the data. Some applications are discussed and proposed. In the end, the forecast comparisons with a linear competitor AR model and non-linear competitor Markov-switching model are conducted, where CLEAR model outperforms the AR model, but does not show advantages over Markov-switching model.