AbstractThis paper argues that the simultaneous use of all leading indicators
may result in the blending of two different sets of information, which
could lead to less accurate predictions of a future recession. We divide
six of Taiwan's leading indicators into two different sectors, the real and
financial sectors, and distinctly demonstrate that the two sectors may very
well reveal different information. Three inconsistent, or even divergent,
movements are found for 1988, 1991 and 1994, implying that the factor
extracted from the real side may be different from that from the financial
side. Thus, in contrast to the one-factor model typically used, we suggest
a two-factor model. We compare four Markov Switching models, and it is
evident that the predicted recessions based on the two-factor one-state
model seem to outperform other models. The second best is the one-factor
model which is only based on the real side variables, followed by the
one-factor model with four variables. The worst model is that which simply
uses financial variables. The results support our argument to use the
two-factor model.