Journal of Economics and Management  
  Volume 15, No. 2  
  August, 2019  

Taiwan Stock Market and the Macroeconomy: A Smooth-Transition Approach


  Christos Michalopoulos  

Department of Economics, Soochow University, Taiwan




We employ a two-regime smooth transition regression model with a logistic transition function to measure the degree of the nonlinear interaction between the Taiwanese stock market and macroeconomic indexes for economic growth (GDP), price stability (CPI), money growth (M2), risk-free rate (Taiwan T-Bills Rate), US exchange rate, and the US stock market’s Dow Jones index. Starting with eight lags for each variable, we employ the LASSO statistical methodology to pick a few significant regressors, and building on the resulting specification, we find strong statistical evidence of nonlinearity, with the Dow Jones index playing the switching variable. The results of the fitted smooth transition model suggest two distinct bull and bear-type regimes for the stock market index with complex, significant, and asymmetric effects due to its lags and other variables. In a simple 4- and 8-step ahead forecasting exercise, our nonlinear model does not seem to outperform the linear specification for most forecasting accuracy measures employed; hence, we are not able to confirm recent results by Guidolin et al. (2014), suggesting that nonlinear models forecast better. Finally, we fit linear and nonlinear specifications by splitting our dataset into two periods, the pre- and post-Great Financial Crisis, as dated by the NBER, and we find intricate nonlinear effects differing between the two periods with the post-GFC fit being better in all statistical measures employed.




Keywords: Nonlinear modeling, smooth transition, LASSO, forecasting. 



JEL classification: C01, C52.