Nonlinearities in the Economic Growth Rates of Taiwan and Hong Kong: A Bayesian Threshold Autoregression Approach

Abstract

We analyze the extent to which low frequency movements in the economic activity of Taiwan
and Hong Kong can be explained by those of their largest trading partners. We estimate
posterior marginal probabilities across 312 different specifications for each country and
computes Bayes' Factors as a model selection mechanism. Data suggests that a best fitting
model requires a different specification for Hong Kong and Taiwan. Results show that
economic growth rates in both Hong Kong and Taiwan are well described by a threshold
model but with different types of nonlinear effects. Hong Kong’s economic growth has a
nonlinear effect on its expected future growth, but is unaffected by growth in the US, Japan
or Taiwan. In contrast, we find a significant nonlinear spillover effect from Japan to Taiwan.