Earticle

Artificial neural networks (ANNs) with the logistic transforms are popular methods to increase the accuracy of performance forecasting due to their functional flexibility. In this paper, we estimate the accuracy of the ANNs models by conducting a data-driven search for optimal specifications. Our tests on foreign exchange rate forecast for the Korean won/US dollar show that the ANNs are superior to linear models. The superiority of the ANNs, however, does not hold for the Japanese yen/US dollar exchange rates. We use the success ratio (SR) as the out-of-sample forecasting performance evaluation. The directional accuracy (DA) test and the forecast comparison statistics of Diebold and Mariano (DM) are applied to assess the relative forecast performance of the ANNs as well. For the Korean won spot rate, it seems that there is much to be gained by using the ANNs for predicting the direction of change. The DA test results also show that the SR from the ANNs is generally greater than that from the AR models. For the Japanese yen, the ANNs achieve a lower SR in most cases. The balance between the in-sample fit and the out-of-sample forecast performance is well achieved for the Korean won exchange rate, but not for the Japanese yen exchange rate. This study is significant in that no previous works that evaluated the accuracy of exchange rate forecasts have applied a variety of tests in order to ascertain whether the ANNs of out-of-sample forecast performance actually achieve directional accuracy as economic criteria.

To examine the common assumption that the monetary fundamentals are stationary, we test the monetary model of exchange rate determination. In specific, we test whether nominal exchange rates are cointegrated with the monetary fundamentals, using data spanning from January 1971 to March 2011. The results show that the Phillips and Ouliaris test fails to reject the null of no cointegration while we reject the null hypothesis of no cointegration with Johansen’s trace test. We also explore the out-of-sample forecasting performance of the long-horizon regression of exchange rate returns on the deviation of the log exchange rate from the monetary fundamentals. For the sample of data, the DOLS estimation does not yield cointegrating coefficient estimates that accord with the theoretical values implied by the monetary model of exchange rate determination. This may bring about critical results asserting that there is no evidence of exchange rate predictability based on the monetary fundamentals in the sample. Despite the use of long span data, no evidence is found in favor of the monetary exchange rate model using the Johansen procedures. We compare the out-of-sample forecasting performance of the monetary model and confirm that forecasting power of the monetary model is insignificant.

In this study we examine whether daily S&P 500 index volatility can be modeled parametrically as a long-memory process by extending an integrated process to a fractionally integrated one. The modified R/S test statistic and others are significant at the 1% level of significance, so we reject the null hypothesis of no long-term dependence. We have found that there is strong evidence for long memory in the series analyzed. We compare the out-of-sample forecasting performance of volatility models from 1962 to 2009. For various forecasting horizons, the long-memory FIGARCH model tends to make more accurate forecasts. Our empirical finding that the index volatility has long memory is consistent with prior evidence showing that an asset market volatility model such as plain GARCH puts too much weight on recent observations in the estimation process relative to those of the past. The forecasting model with the lowest MSFE and VaR forecast error among the models we consider is the FIGARCH model. In terms of forecasting accuracy, it dominates the widely accepted GARCH and rolling window GARCH models. We find that the White’s reality check p-values for the FIGARCH (1, 1) expanding window model reject the hypothesis that there exists a better model than the two benchmark models. The Hansen’s p-values report the same results.