PDF Ebook Exchange Rate Determination of Tl/Us$: A Co-Integration Approach
Determination of exchange rates has been of great concern for policy makers in today’s liberalized and integrated world economies. Examining the course of exchange rates would help researchers conduct empirical investigations for testing the coherence of international macroeconomic theories such as purchasing power parity (PPP) and uncovered interest parity (UIP) as well as theories explaining the determination of exchange rates assuming open economy conditions. Such researches would reveal the extent to which discretionary economic policies can succeed in attaining the ex-ante policy targets and provide knowledge of forecasting performances of different modeling approaches following the developments in estimation techniques as for the future courses of economic variables. Especially, for a small and open developing country such as Turkey policy design process by the policy makers should be inclusive of the stylized facts based on these researches. Kesriyeli (1994), Metin (1994), Telatar and Kazdagli (1998), Bahmani-Oskooee and Kara (2000), Gokcan and Ozmen (2001), Dulger and Cin (2002), Civcir (2003a), Civcir (2003b), Civcir (2003c), Yazgan (2003), Erlat (2003), Ozdemir (2004) and a recent paper by Saatcioglu et al. (2007) give some empirical findings upon these issues of interest for the Turkish economy.
Of all these contemporaneous theoretical developments, a vast literature has been attributed to modeling the behavior of exchange rates so as to see whether monetary fundamentals are able to explain long-run course and short-run dynamics of exchange rates. What is of considerable interest in the economics literature is also to examine how well the out-of-sample forecasts fit to the actual data when assessing various estimation methods for forecasting purposes. Following the seminal paper by Meese and Rogoff (1983) indicating that fundamental based structural models of exchange rate do not beat the performance of the naïve random walk models in out-of-sample forecasts, there has been an extensive controversy upon these issues of interest. Researchers tend to explore whether the models based on structural relations or driven by naïve-random walks or considering more recent multivariate co-integration techniques, both assuming non-stationarity of data in the level form and preserving long-run knowledge of economic relations, must be of special interest and to the extent that they produce more accurate estimates models have been accepted to be superior when compared with the others.
However, some recent literature based on monetary model exchange rate determination yield estimation results supporting fundamental based models in the construction of exchange rates and give evidence in favor of out-of-sample forecasting performances for the fundamental based models against those of the random walk models particularly over relatively long horizons. In this sense, MacDonald and Taylor (1993) give some support to the flexible price monetary model (FPMM) forecasts of Deutsche mark-US dollar exchange rate in a dynamic error correction framework for the 1976-1990 period that are superior to those generated by a random walk forecasting model. McNown and Wallace (1994) search for co-integrating relationships leading to the monetary model of exchange rate determination for Israel, Chile and Argentina experiencing rapid monetary inflations. Their findings using data for the post-1973 period till the late-1980s not only support long-run relations among the variables of the monetary model but are also sensitive to the model specification as for the appropriate signs of the model in the sense that co-integrating vectors with plausible estimates are only obtained for Chile and Argentina but not for Israel. Likewise, Moosa (2000) examining the 1919-1923 German hyperinflation period gives evidence to the monetary model of exchange rate determination. Mark (1995) employing US dollar prices of the Canadian dollar, the Deutsche mark, the Swiss franc and the Japanese yen finds that long-horizon changes in the logarithm of spot exchange rates are predictable. He gives evidence to that the out-of-sample point predictions from fundamental based models generally out-perform the driftless random walk model at the longer horizons. Kilian (1999) and Berkowitz and Giorgianni (2001) present a criticism to Mark’s (1995) methodoloy dealing with the data generating process used for predictability. Chinn and Meese (1995) support Mark’s conclusions to a great extent and find that fundamental-based error correction models out-perform the random walk model for long-term prediction horizons. Neely and Sarno (2002) give a brief summary of the seminal papers of Meese and Rogoff (1983) and Mark (1995) upon exchange rate determination mechanism considered in the economics literature.
MacDonald and Marsh (1997) also indicate that fully dynamic out-of-sample forecasts from simultaneous equations models incorporating meaningful long-run equilibrium and short-run dynamic relationships are cabaple of significantly out-performing those of a random walk model considering US, Japan, UK and German data for the 1974-1990 period. Cheung and Chinn (1998) apply to a methodology using some consistency tests of evaluating exchange rate forecast rationality for Japanese yen, German Mark and Canadian dollar exchange rates against the US dollar for 1983-1993 and 1987-1993 periods, for which consistency requires that the forecast and the actual series (i) be the same order of integration, (ii) be cointegrated and (iii) yield co-integrating vector consistent with long-run unitary elasticity of expectations. They indicate that the first requirement generally holds, however co-integration fails to hold the longer the horizon. Of the co-integrating pairs, besides, the third requirement is not generally rejected.
Following recent developments in the methodology of estimation techniques, researchers tend to use panel estimation techniques and so examine whether results from panel based studies are able to yield more powerful results in favor of fundamental based models in the long run. In this sense, Groen (2000) employs data for 14 bilateral exchange rates and monetary fundamentals with respect to either the US dollar or Deutsche mark for the 1973-1994 period. However co-integration tests on the time series of individual countries lack of giving evidence for the monetary model, he finds that panel based testing both produces more powerful results supporting monetary model and indicates co-integrating relation between exchange rates and monetary fundamentals. Mark and Sul (2001) examine the long-run relationship between nominal exchange rates and monetary fundamentals in a panel of 19 countries for the 1973-1997 period. They estimate that co-integration between exchange rates and long run fundamentals predicted by economics theory is generally approved by the data and that panel based forecasts indicate significant predictive power of monetary fundamentals for future exchange rate movements. Using the data set of Mark and Sul (2001), Rapach and Wohar (2004) test the long run monetary model of exchange rate determination as well. They first support the evidence that country-by-country estimates of co-integrating coefficients diverge widely from the values predicted by the monetary model and give little evidence for the co-integration between nominal exchange rates and monetary fundamentals in the floating period. But when they apply to the panel estimates of co-integrating coefficients they obtain supportive results in line with the monetary model as for the signs and magnitudes. However, when they analyze the cross-country homogeneity restrictions in the panel estimation procedure they reveal that such assumptions are not supported by the data. Rapach and Wohar (2002) emphasize that results in Groen (2000) and Mark and Sul (2001) should be reexamined for robustness to various subpanels and to formally test for heterogeneity across panel members.
Besides, some recent papers find evidence for the fudamental based models. Abhyankar et al. (2005) using data for the US, Canada, Japan and the UK for the modern floating exchange rate period lend some support to the predictive ability of the exchange rate monetary fundamentals model. Their results indicate that the gain from using the information in fundamentals in order to predict the exchange rate out-of-sample is found substantial. Karfakis (2006) using US and euro area observations for the 1999-2004 period reveals that the exchange rate is co-integrated with money and income differentials and that estimated error correction model out-performs the random walk forecasts. Also Nwafor (2006) supports flexible price monetary model by estimating a long-run equilibrium relationship between Nigerian naira and the US dollar.
In our paper, our aim is to examine the empirical validity of monetary model of exchange rate determination for the Turkish economy and to compare the out-of-sample forecasting performances of the results with those of a naïve random walk model. For this purpose, the outline of the paper is as follows. We first highlight the construction of a simple flexible price monetary exchange rate model as well as some extensions examined in the economics literature. Then, an empirical model for the Turkish economy is constructed which also assesses the out-of sample forecasting performance against naïve random walk model. Finally, the last section summarizes results and concludes.
Download
PDF Ebook Exchange Rate Determination of Tl/Us$: A Co-Integration Approach
Posted in :
