Return to search

Modeling Target Zone with nonlinear regression-the cases of German, Italy and France

The exchange rate target zone has been paid much attention in the early 1990 initially by Krugman (1991).It expressed when exchange rate surpasses the band of exchange rate that implicitly or explicitly determined by the central bank, the central Bank will intervene the foreign exchange by buying or selling foreign exchange to ensure the exchange rate staying inside the band, otherwise, the exchange rate will be allowed to fluctuate inside the band freely.According to Krugman (1991), when economic system faces random disturbances, the exchange rate target zone regime is helpful to narrow down the exchange rate volatility contrast to that in the floating exchange rate regime. That is, the exchange rate target zone has more essential stability,which is called ``honeymoon effect".
In recent decade, Krugman's exchange rate target zone model has been tested empirically.In this thesis, the smooth transition autoregression with target zone (STARTZ) proposed originally by Lundbergh and Ter"{a}svirta (2006) and logistic smooth transition regression with two thresholds (LSTR2) are used to make comparisons for in-sample fitness and out-of-sample forcastability.Furthermore, we also test two important assumptions of the exchange rate target zone model: the credibility assumption and marginal interventions.
The data are constructed with 755 daily spot exchange rates, denominated in Eurpean Currency Unit (ECU), from January 14, 1987 to December 29, 1989, in German, France, and Italy.We split the sample into in-sample (570 observations), and out-of-sample (185 observations), and make use of STARTZ-GARCH and LSTR2-STGARCH to fit the in-sample regimes, and apply Rapach and Wohard (2006)'s Bootstapping to generate the out-of-sample forecasts.
Finally,we make use of Diebold and Mariano (1995)'s predictive accuracy tests to compare the out-of-sample forecastability between STARTZ and LSTR2 models.According to the empirical results, we can find that LSTR2 model has not bad performance in fitting the in-sample and forecasting the out-of-sample data compared to STARTZ model.

Identiferoai:union.ndltd.org:NSYSU/oai:NSYSU:etd-0730107-144100
Date30 July 2007
CreatorsTsai, Shang-ying
ContributorsChingnun Lee, Yung-hsiang Ying, Ming-Jang WENG
PublisherNSYSU
Source SetsNSYSU Electronic Thesis and Dissertation Archive
LanguageCholon
Detected LanguageEnglish
Typetext
Formatapplication/pdf
Sourcehttp://etd.lib.nsysu.edu.tw/ETD-db/ETD-search/view_etd?URN=etd-0730107-144100
Rightsunrestricted, Copyright information available at source archive

Page generated in 0.0014 seconds