Exchange Rate Series Research Articles

Forecasting Daily and Monthly Exchange Rates with Machine Learning Techniques

AbstractIn this paper we propose and test a forecasting model on monthly and daily spot prices of five selected exchange rates. In doing so, we combine a novel smoothing technique (initially applied in signal processing) with a variable selection methodology and two regression estimation methodologies from the field of machine learning (ML). After the decomposition of the original exchange rate series using an ensemble empirical mode decomposition (EEMD) method into a smoothed and a fluctuation component, multivariate adaptive regression splines (MARS) are used to select the most appropriate variable set from a large set of explanatory variables that we collected. The selected variables are then fed into two distinctive support vector machines (SVR) models that produce one‐period‐ahead forecasts for the two components. Neural networks (NN) are also considered as an alternative to SVR. The sum of the two forecast components is the final forecast of the proposed scheme. We show that the above implementation exhibits a superior in‐sample and out‐of‐sample forecasting ability when compared to alternative forecasting models. The empirical results provide evidence against the efficient market hypothesis for the selected foreign exchange markets. Copyright © 2015 John Wiley & Sons, Ltd.

Nonlinear Taylor rules in Central Eastern European countries

This study applies sequential panel selection method (SPSM), proposed by Chortareas and Kapetanios (2009), to investigate to test the validity of Taylor rules to assess the nonstationary properties of the convergence of the real exchange rates for 10 Central Eastern European countries. The SPSM can be used to decompose a panel of real exchange rate series into two groups: a group of stationary series and a group of nonstationary series. We identify the stationary processes in the panel and demonstrate that Taylor rules holds for 7 of the 10 countries studied. These results imply that the choices and effectiveness of the monetary policies in Central Eastern European economies are highly influenced by external factors originating from the United States. Additionally, our findings highlight that their real exchange rate convergence is a mean reversion toward equilibrium values of Taylor rules in a nonlinear manner.

Exchange Rate Predictability and the Skewness of Bid-Ask Spreads

I find, regardless of the exchange rate series utilized, that the incorporation of conditional (predicted) skewness as an exogenous variable in ARCH models of the exchange rate process induces significant decreases in the price of exchange rate volatility risk alongside decreases in the risk that dealers will incur losses in transactions with informed traders. In out-of-sample tests, I find ARCH models that incorporate the conditional skewness factor (augmented ARCH models) have better forecast power in relation to ARCH models that exclude the conditional skewness factor (non-augmented ARCH models) only within a low risk foreign exchange market. In aggregate, empirical findings provide evidence of an inverse relation between market risk and market efficiency within the cross-section of foreign exchange markets. This inverse relation is not predicted by theories of market efficiency or theories of intertemporal risk-return trade-offs, but can be induced by market frictions within foreign exchange markets.

Forecasting exchange rate using deep belief networks and conjugate gradient method

Forecasting exchange rates is an important financial problem. In this paper, an improved deep belief network (DBN) is proposed for forecasting exchange rates. By using continuous restricted Boltzmann machines (CRBMs) to construct a DBN, we update the classical DBN to model continuous data. The structure of DBN is optimally determined through experiments for application in exchange rates forecasting. Also, conjugate gradient method is applied to accelerate the learning for DBN. In the experiments, three exchange rate series are tested and six evaluation criteria are adopted to evaluate the performance of the proposed method. Comparison with typical forecasting methods such as feed forward neural network (FFNN) shows that the proposed method is applicable to the prediction of foreign exchange rate and works better than traditional methods.

Improving predictability of time series using maximum entropy methods

We discuss how maximum entropy methods may be applied to the reconstruction of Markov processes underlying empirical time series and compare this approach to usual frequency sampling. It is shown that, in low dimension, there exists a subset of the space of stochastic matrices for which the MaxEnt method is more efficient than sampling, in the sense that shorter historical samples have to be considered to reach the same accuracy. Considering short samples is of particular interest when modelling smoothly non-stationary processes, which provides, under some conditions, a powerful forecasting tool. The method is illustrated for a discretized empirical series of exchange rates.

Are real exchanges rate series really persistent?: evidence from three commonwealth of independent states countries

This article sets out to examine the degree of persistence in the real exchange rates of Kazakhstan, Kyrgyzstan and Russia, and tests the validity of Purchasing power parity (PPP) using monthly data covering 1995:01–2013:12 period. The sum of autoregressive (AR) coefficients is used in order to examine persistence of the real exchange rate series and grid-bootstrap method is employed for the confidence intervals. The tests performed suggest two results: (1) Covering the full sample and sub-sample periods, excluding Kyrgyzstan in 1995:01–1998:07 and Russia in 2008:09–2013:12 periods, disregarding the structural breaks in the data generating process, there is high persistency in real exchange rates; (2) there is evidence in support of PPP covering the full sample for every country except for Kyrgyzstan in 1995:01–1998:07 and Russia in 2008:09–2013:12 periods.

Revisiting Purchasing Power Parity in OECD

This study revisits Purchasing Power Parity theory (PPP) in the 34 OECD countries during January 1994–August 2013. We use a new panel stationary test with both sharp breaks and smooth shifts, a novel approach to panel unit-root testing, proposed by Bahmani–Oskooee et al. (2014). The results indicate that the PPP holds in half of the 34 OECD countries. These results indicate the importance of proper modelling of both sharp breaks and smooth shifts in real effective exchange rate series of OECD countries.

Kırılgan Beşlide Satın Alma Gücü Paritesi (SAGP) Hipotezinin Test Edilmesi

Due to price stability has been a privileged macro-economic target, the validity of purchasing power parity (PPP) hypothesis attracted more attentions in recent studies. The aim of this study is to examine the validity of purchasing power parity hypothesis for emerging markets (India, Brazil, Indonesia, Turkey, and South Africa) which are called “fragile five” by using the method of co-integration in the the period of 2003:12013:8. Then, the data covers monthly series of nominal exchange rates and consumer price indices for fragile five and consumer price index for the US. Findings of the study illustrated that there is a long term relationship between the nominal exchange rates and consumer price indices of Brazil, India, South Africa, Turkey and consumer price indices of the US. Thus the linear integration of the series are stationary and strongly validity of purchasing power parity is current. However, results of the analysis points out an exception for Indonesia that the long term relationship between the variables does not exist and series are not co-integrated. So, the hypothesis of purchasing power parity is rejected for Indonesia. It can be argued that the Central Banks of Fragile five countries are not capable of coping with risks crated by fluctuations in the real exchange rates, because they implement inflation targeting regime. As a result, the fluctuations of the real exchange rate around its long term equilibrium value has crucial importance for the central banks’ political sovereignty

Modeling Exchange Rate Dynamics in Egypt: Observed and Unobserved Volatility

The underlying study focuses on estimating and forecasting the volatility of exchange rate in Egypt based on ARCH type models and the State Space (SS) models, namely; the Stochastic Volatility (SV) and the Time-Varying Parameter (TVP) models. Moreover, the paper tests the predictive power of the conducted models to come up with a powerful technique that gives the best forward-looking stance of the exchange rate. Empirically, the paper utilizes daily exchange rate data spanning from January 2003 till June 2013. Evidently, it is found that the exchange rate returns in Egypt suffer from the volatility clustering phenomenon and that there exists a time-varying variance in the exchange rate series that has to be appropriately dealt with, while modelling nominal exchange rates. Additionally, with regard to the link between the volatility occurring in the stock market in Egypt and the volatility of the exchange rate market, it is found that there is a risk mismatch between the two markets. Therefore, further research is recommended in the future to suggest other exogenous variables that can help in explaining the volatility in the exchange rate returns in Egypt.

Oil Price and Exchange Rate in Malaysia: A Time-Frequency Analysis

The study analyzed the Granger-causal relationship in the time-frequency framework between return series of real oil price (ROP) and real effective exchange rate (REER) for Malaysia. In doing so, study relied on time-frequency framework of the Granger-causality, which is based on continuous wavelet approach. We found that the causal and reverse causal relations between oil price and real exchange rate vary across scale and period viz., during late 1989, in the time scale of 8~10 months, both variables were in phase and ROP was leading and both variables were out of phase and ROP was leading (a) in 1990-1991, in the time scale of 12~16 months, (b) in 1997 -1998 in the time scale of 10~16 months, (c) in 2001-2003, in time scale of 9~15 months, and (d) in 2005 and early 2006, in the time scale of 2~7 months. Further, evidence shows that during 1989-1998, in 32~48 months scales, variable were in phase and ROP was lagging and throughout the study period, in 60~64 months scale, variables were in phase and ROP was leading. Hence, our evidence show that there is evidence of both cyclical and anti-cyclical relationship between ROP and REER at shorter time scales however, throughout study for higher scales REER was lagging and receiving cyclical effects of ROP shocks. Findings obtained in the study have implications for central monetary authority of Malaysia in the formulations of appropriate monetary and exchange rate policies and for traders in the formulations of effective risk management.

The relationship between Brazilian Real and the US Dollar after corporate governance

The purpose of this study is to determine an econometric model forecasting the nominal exchange rate between the Brazilian Real and the US Dollar. Monthly data from January, 2005 through February, 2011 is used. The Dickey-Fuller (DF) and augmented Dickey-Fuller (ADF) for the existence of a unitary root in the series of the nominal exchange rate (R$/U$), the inflation differential between Brazil and the USA, as well as the interest rate differential between Brazil and the USA, which are integrated in first order, or I(1) are used. The econometric estimate is based on a methodology which considers the specification of a general model to a more specific one. According to the chow test, the forecast for six months is stable, indicating exante coefficients in a determined period of significance.

Natural Convection in Near Shore Waters Subject to Periodic Thermal Forcing

The variation of water depth across shore in nearshore waters results in differential heating which generates natural convection. Field observations have demonstrated the significance of this buoyancy-driven flow in promoting exchanges across shore. The present investigation is concerned with natural convection in nearshore waters subject to periodic thermal forcing at the water surface. Flow response at different stages of the periodic forcing has been illustrated by isotherms and streamlines from numerical simulations, and an evident effect of phase delay is observed. Analysis on the time series of horizontal exchange rate reveals that the phase delay of flow response decreases with increasing period. For sufficiently large period, the heating and cooling phase is characterized by low and high values of horizontal exchange rate respectively.

Purchasing Power Parity in the SAARC Region: Evidence from Unit Root Test with Cross-Sectional Dependence

Mixed results on the validity of Purchasing Power Parity (PPP) relationship in South Asian countries motivates this paper to conduct a further inquiry. Existing studies lack appropriate treatment of cross-country dependence in the testing procedure. In this paper we employ a method that allows us to identify the degree of cross-sectional dependence (CSD) and apply panel unit root test to accommodate this dependence on the real exchange rate series of five South Asian countries in the South Asian Association of Regional Cooperation (SAARC) which include Bangladesh, India, Pakistan, Sri Lanka and Nepal. We find evidence of strong cross-country dependence, cross-country correlation being 0.735. Our panel unit test results support the validity of long-run PPP in the sample countries. This result is in contrast to the previous studies in similar countries which did not accommodate CSD in their estimation. This finding implies that real shocks do not have any permanent effect on the real exchange rate and other things remaining the same, no active policy intervention is warranted for the sustainability of external balance.

A t-distribution based particle filter for univariate and multivariate stochastic volatility models

Stochastic Volatility (SV) model usually assumes that the distribution of asset returns conditional on the latent volatility is normal. Previous approaches to estimation of SV model have mostly focused on Gaussian filters in practice. This paper analyzes SV model with the student-t distribution and compares the distribution with mixture-of-normal distributions of Kim and Stoffer [22]. A Sequential Monte Carlo with Expectation–Maximization (SMCEM) technique based on student-t distribution is developed to estimate the parameters for the extended volatility model. The SMC method, or particle filter based on student-t distribution, which is heavier tailed than Gaussians, provides an approximate solution to non-Gaussian estimation problem and hence more robust. Our empirical analysis indicates that extension of the SV model such as a specification of the error term with student-t distribution in the return equation dominates the normal mixture distribution. Additionally, the t-distribution based particle filter is applied to a multivariate stochastic volatility model. It is again shown that the student-t based algorithm performs quite well in explaining the joint dynamics in the volatility of a set of four exchange rates series.

The efficacy of intervention in a chaotic foreign exchange market: an empirical study in INR-USD exchange rate series

The traditional method of containing forex volatility is by proper controlling of domestic interest rate and foreign exchange reserves. Such instruments, however, have often failed raising doubts about the efficacy of central bank intervention in forex markets. Recent literature tries to rescue the policy makers by pointing toward the intrinsic dynamics or more specifically, the possible chaotic nature of forex markets and its policy implications. Since volatility is endogenous and no long-run predictions are possible in a chaotic market, traditional methods of interventions are likely to fail. This paper explores this issue in context of the Rupee-Dollar exchange rate. It finds that the central bank intervention to be indeed ineffective in an inherently chaotic market. Thus, inefficacy of intervention in forex market could possibly be explained in terms of its intrinsic dynamics. These findings call for a changed policy outlook that could consider the non-linear nature of the forex rates to successfully control its volatility.

Developing of approaches for the unhomogeneus time series analysis based on statistical characteristics

In this paper, a study of non-homogeneous time series, as a special case of a class of non-stationary time series, which are not given to stationary by integrating the 1st and 2nd order was carried out. Because of the presence of non-homogeneous components, time series is characterized by a non-linear trend, periodic components with variable amplitude, as well as a non-permanent structure. A method for identifying the series of this class based on statistical characteristics, using statistical tests and criteria was proposed. This method allows reasonable approach to selecting a method for forecasting a number of changes with respect to all the hidden features of the original data, as well as the reasons and assumptions of various forecasting methods. For example, in the construction of forecasting models of the non-homogeneous time series using statistical forecasting methods, such as AR (p), MA (q), ARMA (p, q), AR (p) + linear trend, ARCH (p), GARCH (p, q), SV (p), is inadequate in view of violating the requirements for applying these methods, in particular time invariance. An adaptive approach for forecasting non-homogeneous time series based on the method of singular spectrum analysis and presentation of time series in several phase spaces is proposed in the paper. Another problem in forecasting the time series is a risk assessment in decision-making based on the forecasting model. A method for estimating the maximum level of losses in forecasting non-homogeneous time series with a given probability was developed in the paper. A confidence interval of forecasting non-homogeneous time series of exchange rates, which allows estimating the adoption of one or other solutions depending on the amount of the transaction was built. The results obtained in this paper can be used for a sound approach to the selection of a forecasting method in automated systems, as well as in the evaluation of making management decisions.

Revisiting purchasing power parity in major oil-exporting countries

Univariate unit-root tests of the purchasing power parity (PPP) are said to suffer from low power. Following the literature, we apply the sequential panel selection method combined with a Fourier function to test the PPP in six major oil exporting countries. The results support PPP in all six countries except Russia. Our results point to the importance of proper modelling of structural breaks and non-linearities in the real exchange rate series of these countries.

Currency Crisis Prediction Using Wavelet-Based Support Vector Machine

A currency crisis is typically a kind of rare event, but it hurts sustainable economic development when it occurs. A novel method of wavelet-based support vector machine (SVM) is proposed to predict financial crisis events for early-warning purposes in this paper. In the proposed method, currency exchange rate, a typical currency indicator that usually reflects economic fluctuations, is first chosen. Then a wavelet decomposition algorithm is applied to the currency exchange rate series. Using the wavelet decomposition procedure, some details and features of the currency exchange rate series, with different scales, can be obtained. Using these details and features, a wavelet-based SVM learning paradigm is used to predict future currency crisis events, based upon some historical data. For illustration purpose, the proposed wavelet-based SVM learning paradigm is applied to exchange rate data of two Asian countries to evaluate the state of currency crisis. Experimental results reveal that the proposed wavelet-based SVM learning paradigm can significantly improve the generalization performance relative to some popular forecasting methods.

Can a simple structural time series model beat the random walk?

This paper tries to address the question that if the long run PPP holds, then there should exist a structural model which can outperform the random walk in out of sample forecasting. We propose an ARFIMA based model with log of the independent variable as an explanatory variable and make a comparison study of this structural model with the benchmark random walk model. Then, we compare our results with that as obtained by Engel and Hamilton, and by Clarida, Sarno, Taylor and Valente. We present the standard ARFIMA model and show how can make an extension of it so that it becomes a variant of ARFIMA and name it as YQ-ARFIMA, then construct a bivariate model relating the dependent variable yt and ln yt , and with that, perform an impulse response function analysis of the predictive ability of ln yt . We also transform the YQ-ARFIMA into a moving average representation, and thereafter perform the impulse response function analysis again, then make a comparison study between the standard ARFIMA and the YQARFIMA by comparing the out of sample forecasting ability of each one of them with the benchmark random walk model. After that, compare the performance of YQ-ARFIMA with that of the Markov switching model put forward by Engel and Hamilton, and the MSIH(3)-VECM as put forward by CSTV. Last, we test the robustness of the YQ-ARFIMA by fitting it into different exchange rate series spanning the five continents of the globe, then, test the consistency of the forecast by YQ-ARFIMA by a cointegration technique. By using the loss functions RMSE and MAPE, cointegration consistency in forecasts and impulse response function analysis, we have shown beyond doubt that theYQ-ARFIMA model is very much superior in forecasting ability.

Reexamining the PPP hypothesis: A nonlinear asymmetric heterogeneous panel unit root test

In this study, we re-examine the PPP hypothesis in the light of the new developments in the unit root testing literature. The recent theoretical findings have pointed out that the real exchange rate series exhibit asymmetric nonlinear behavior. A unit root test applied to analyze the PPP hypothesis therefore, should also take into account this asymmetry inherent in the real exchange rate. Different unit root tests that consider the presence of these data features have been developed in the time series literature. However, a true attempt to test the PPP hypothesis should take a panel data approach. To this end, we propose a nonlinear heterogeneous panel unit root test where the alternative hypothesis allows for symmetric or asymmetric exponential smooth transition autoregressive nonlinearity and provide its finite sample properties. We apply our test to the real exchange rates of the 15 European Union countries against the US dollar. While the results of the linear and symmetric nonlinear heterogeneous panel unit root tests are against the PPP hypothesis, the asymmetric nonlinear heterogeneous panel test that we propose gives support for the PPP hypothesis as expected. Therefore, the conclusions drawn from the linear panel unit root tests or the nonlinear panel unit root tests that do not take asymmetry into account might be misleading.