The period of financial crisis (2007/4-2012/6) was a very specific period that the government’s policies were significant matter to exchange rate. This is also the merit of Taylor rule models because Taylor rule models could capture both interest rates spread and output gap. That’s the rationale that the best performance during the financial crisis in this thesis is the Taylor Fundamental model with output gap and FCI spread. This model also performed better than it did in non-financial crisis period.
After the financial crisis (2012/6-2014/8), the global market was in recovery. The GDP of both Taiwan and US have become more and more important, so that output gaps still a variable that could not be ignored. Taylor rule models still stand an important role in predicting exchange rate. Moreover, Taylor differential models performed better than Taylor Fundamental models. The best performance during the non-financial crisis period in this thesis is the Taylor rule differential models with output gap.
About the unemployment gap, although Molodtsova and Papell (2012) found it was very successful using Taylor rule models with unemployment gap for the exchange rate of US dollar per Euro, it was certainly not suitable for USD/NTD in Taiwan. Taylor rule models with output gap were still performed better during both financial crisis period and non-financial crisis period in Taiwan.
29
Based on the results we found in this thesis, as we could see in Table 10, the best strategy is that we should use Taylor Fundamental model during specific period like financial crisis. During the periods without special financial events, we should use Taylor differential models. If we would like to predict a long period of time, which has a high probability of including structure changing, we should use Purchasing Power Parity model.
In summary, Taylor rule models with output gap are the most useful models in Taiwan for predicting exchange rate of USD/NTD, but we still could not forget Purchasing Power Parity model. It also gave us a stable outperformed predictability during the whole period.
30
Table 4 Summary of Regressions
Notes: the table shows the regression of each model. This table will be used in the following table.
Model Regression
31
Table 5 Summary of coefficients (Regression sample: 1996/1-2007/3)
Notes: the first column for Taylor differential models represents the coefficient of 𝜔𝑖. The numbers in the bracket are the t-statistics of each coefficient. *, **
and *** denote the test statistics significant at 10%, 5% and 1% level, respectively.
(𝜋𝑡− 𝜋𝑡∗) or 𝜔𝑖 (𝑦𝑡− 𝑦𝑡∗) or (𝑢𝑡− 𝑢𝑡∗) (𝑠𝑡− 𝑠𝑡∗)
32
Table 6 Summary of coefficients (Regression sample: 1996/1-2012/6)
Notes: the first column for Taylor differential models represents the coefficient of 𝜔𝑖. The numbers in the bracket are the t-statistics of each coefficient. *, **
and *** denote the test statistics significant at 10%, 5% and 1% level, respectively.
(𝜋𝑡− 𝜋𝑡∗) or 𝜔𝑖 (𝑦𝑡− 𝑦𝑡∗) or (𝑢𝑡− 𝑢𝑡∗) (𝑠𝑡− 𝑠𝑡∗)
33
Table 7 Summary of performance (Duration 2007/4-2014/8)
Notes: the table reports the out of sample 𝑅𝑜𝑜𝑠2 , ∆𝑅𝑀𝑆𝐸 and CW statistic of each model. * denotes test statistics that performed better than random walk model for 𝑅𝑜𝑜𝑠2 and ∆𝑅𝑀𝑆𝐸. For CW test statistics, * denote test statistics are significant at
34
Table 8 Summary of performance (Duration 2007/4-2012/6)
Notes: the table reports the out of sample 𝑅𝑜𝑜𝑠2 , ∆𝑅𝑀𝑆𝐸 and CW statistic of each model. * denotes test statistics that performed better than random walk model for 𝑅𝑜𝑜𝑠2 and ∆𝑅𝑀𝑆𝐸. For CW test statistics, * denote test statistics are significant at
35
Table 9 Summary of performance (Duration 2012/7-2014/8)
Notes: the table reports the out of sample 𝑅𝑜𝑜𝑠2 , ∆𝑅𝑀𝑆𝐸 and CW statistic of each model. * denotes test statistics that performed better than random walk model for 𝑅𝑜𝑜𝑠2 and ∆𝑅𝑀𝑆𝐸. For CW test statistics, * denote test statistics are significant at
36
Table 10 Summary of performance ranking using 𝑅𝑜𝑜𝑠2
Notes: The number stand for the ranking of 11 models during each period. * denotes test statistics that performed better than random walk model for 𝑅𝑜𝑜𝑠2 and ∆𝑅𝑀𝑆𝐸
37
Reference
1. Blinder, Alan S. and Ricardo Reis (2005), “Understanding the Greenspan Standard”, The Greenspan Era: Lessons for the Future, Federal Reserve Bank of Kansas City, 2005, pp. 11-96.
2. Chinn, Menzie D. (2006), “A Primer on Real Effective Exchange Rates:
Determinants, Overvaluation, Trade Flows and Competitive Devaluation”, Open economies review 17.pp. 115–143.
3. Campbell, John Y., and Thompson, Samuel B. (2008), “Predicting Excess Stock Returns Out of Sample: Can Anything Beat the Historical Average?” , Review of Financial Studies. 21(4).pp. 1509-1531..
4. Clark, Todd E., and West, Kenneth D. (2006), “Approximately normal tests for equal predictive accuracy in nested models”, NBER Technical Working Paper No.
326.
5. Clark, Todd E., and West Kenneth D. (2006), “Using out-of-sample mean squared prediction errors to test the martingale difference hypothesis”, Journal of Econometrics 135, pp. 155-186.
6. Diebold, Francis X. , and Mariano ,Roberto S. (1995), “Comparing Predictive Accuracy”, Journal of Business and Economic Statistics 13, pp. 253-265.
7. Diebold, Francis X. (2013), “Comparing Predictive Accuracy, Twenty Years Later:
A Personal Perspective on the Use and Abuse of Diebold-Mariano Tests”, Journal of Business and Economic Statistics, Published online: 26 Jan 2015
8. Engel, Charles, Nelson C. Mark, and Kenneth D. West (2008), “Exchange Rate Models Are Not as Bad as You Think”, NBER Macroeconomics Annual 2007, University of Chicago Press, pp. 381-441
9. Mark, Nelson (1995), “Exchange Rate and Fundamentals: Evidence on Long-Horizon Predictability”, American Economic Review 85, pp. 201-218.
38
10. Meese, Richard A. and Kenneth Rogoff (1983), “Empirical Exchange Rate Models of the Seventies: Do They Fit Out of Sample?”, Journal of International Economics 14, pp.3-24.
11. Molodtsova, Tanya, and David H. Papell, David (2009),”Exchange Rate Predictability with Taylor Rule Fundamentals”, Journal of International Economics 77, pp.167-180.
12. Molodtsova, Tanya, and David H. Papell (2012), “Taylor Rule Exchange Rate Forecasting During The Financial Crisis”, NBER Working Paper No. 18330.
13. Nikolsko-Rzhevskyy, Alex, and Papell, David H. (2012), “Taylor’s Rule versus Taylor Rules”, International Finance Volume 16, Issue 1, pp. 71–93.
14. Pasquale Della Corte, and Ilias Tsiakas (2012), “Statistical and Economic Methods for Evaluating Exchange Rate Predictability”, Handbook of exchange rates 2012.
15. Rudebusch, Glenn (2010), “The Fed’s Exit Strategy for Monetary Policy”, Federal Reserve Bank of San Francisco Economic Letter 18, pp. 1-5.
16. Taylor, John B. (1993), “Discretion versus policy rules in practice”, Carnegie-Rochester conference series on public policy 39, pp. 195-214.
17. Woodford, Michael (2001), “The Taylor Rule and Optimal Monetary Policy”, The American Economic Review. pp. 232-237.
18. Welch, Ivo and Goyal, Amit (2008), “A Comprehensive Look at The Empirical Performance of Equity Premium Prediction”. Review of Financial Studies.
21(4).pp. 1455-1508.