Unveiling the Time-dependent Dynamics between Oil Prices and Exchange Rates: A Wavelet-based Panel Analysis.

AuthorKarlsson, Hyunjoo Kim
  1. INTRODUCTION

    Since the pioneering works of Golub (1983) and Krugman (1983), the nexus of oil prices and exchange rates has drawn not only academic interest but also the attention of policy makers. The world-market price for crude oil is one of the most important macroeconomic variables for both oil-exporting and importing economies, and it has far-reaching effects not only for currency values, but also for inflation, current accounts, and investors' behavior. Especially for oil-exporting countries, there has been significant media coverage of the link between oil price fluctuations and exchange rates. (1) Since the U.S. dollar is the major invoicing currency of international oil markets, numerous studies on this topic have documented the relationship between oil prices and dollar exchange rates. A weak dollar may keep oil-exporting countries vigilant because the revenues from oil exports may be jeopardized. By contrast, oil-importing countries may benefit from a weak dollar because a given volume of oil would cost less in their own currencies. Although the relationship between oil prices and exchange rates is firmly placed from a theoretical perspective, the empirical results in the existing literature seem to vary depending on whether the dollar movements are vis-a-vis currencies of the other net oil exporters (documenting a dollar depreciation with an oil price increase) or against net oil importers (documenting a dollar appreciation with an oil price increase) (Lizardo and Mollick, 2010; Reboredo, 2012). (2) The signs of the relationships seem to disagree with different exchange rates--nominal and real--that are available at different frequencies of observations--daily and monthly observations, respectively (Ferrao, Rogoff and Rossi, 2015). Due to media coverage and everyday foreign exchange situations, most of us are more acquainted with the nominal exchange rate than the real exchange rate. However, what is usually relevant for households or companies is the goods or services that can be bought in exchange for a certain amount of U.S. dollars, euros, or any other currencies at their disposal. These properties are measured by real exchange rates. Therefore, when investigating the relationship between oil prices and exchange rates, the current paper focuses on the movement of the real effective exchange rates (REER). (3)

    The main objective of this paper is to re-examine the relationship between oil prices and real exchange rates following Chen and Chen (2007). Thus, this paper analyzes the link between oil prices and real exchange rates and builds on previous studies, such as Zhou (1995), that find that oil price fluctuations play a major role in explaining real exchange rate movements. The countries included in the study are major oil exporters with floating exchange rates, with or without inflation targeting. There are two convincing reasons that justify the need for empirical analysis of oil prices and exchange rates of these major oil exporters (with similar exchange rate regimes). First, as mentioned above, how the currency value of a country changes depends on the country's profile as an oil exporter or importer. Therefore, exploring the relationship between oil prices and exchange rates focusing on major oil exporters may show stronger statistical evidence of the relationship. Second, by including major oil-exporting countries with floating exchange rates and inflation targeting, we can draw policy implications for central banks.

    In addressing the issues above, this paper employs wavelet time-scale analysis for investigating the dynamic relationship between oil prices and exchange rates. The difference between using wavelet analysis and traditional methods of studying short- and long-run relationships (error-correction models and the cointegration relationship, respectively) is that, by using wavelets, we can avoid losing relevant variable information by using first differences. Wavelet analysis can provide a broader spectrum of relationships at different time scales than traditional approaches. There are two main versions of wavelet analysis: the discrete and continuous wavelet transform (DWT and CWT, respectively). Both methods have their advantages and disadvantages compared to each other. One of the advantages of CWT is that it gives the researcher more freedom regarding the choice of wavelet filters compared to its discrete counterpart. (4) Using CWT is advantageous especially when studying transient relationships, business-cycle synchronizations, and lead and lag structures between economic variables. (5) In comparison with CWT, one of the main advantages of DWT is that it is possible to decompose a variable into a trend, cycle, and noise in a way that is very similar to the Baxter and King (1999) band-pass filter, which has contributed to its popularity in economics (Yogo, 2008). The method used in the current paper is a version of the DWT denoted as the maximum overlap discrete wavelet transformation (MODWT), which is applied in Ramsey and Lampart (1998) and Gallegati et al. (2016), among others. MODWT shares the above-mentioned advantages of DWT and, according to Aguiar-Conraria and Soares (2014), it is the most common method in economic analysis, since the selection of time parameters is not sparse using MODWT. In contrast to most other methods, MODWT is comparable to traditional time-series regression settings. For instance, it can be applied for linear regression, causality analysis, et cetera, so the MODWT is useful for directly analyzing the signs and magnitudes of the regression coefficients at different time scales. Furthermore, it also enables us to use panel methods as proposed by Gallegati et al. (2016), which still has not been developed for CWT. Due to the reasons described above, it is evident that the MODWT is more suitable than CWT for this type of data and research purpose.

    Our empirical analysis reveals that traditional univariate and panel methods, such as the Engle and Granger (1987) approach, as well as a modern and more powerful panel method based on this approach suggested by Pedroni (2004), fail to detect a relationship between oil prices and exchange rates at any conventionally used significance levels. The new wavelet panel method is thus used as a more powerful alternative estimator in the situation of non-cointegrated panels, which is a distinguishing feature of this paper compared to previous studies. Using this novel approach, the results clearly indicate a positive relationship for the major oil exporters' real oil prices and REER. Both in terms of magnitude and in terms of significance, this relationship is more clearly exhibited at the larger time scales of 4-8 and 8-16 months. The relationship for the time series are not as consistently significant for the smaller time scales of 1-2 and 2-4 months. Additionally, at these smaller time scales, the individual coefficient estimates exhibit different signs, which is an indication that merging these time series into a panel may risk causing a misspecification. Note that, in comparison to larger time scales, the panel of the coefficients are rather close to zero in magnitude. The time-dependent dynamics of the relationships between real oil prices and REER that are found by using wavelet analysis suggest that the aggregation of all time scales in traditional methods might lead to blunt econometric analyses, thereby resulting in type 2 errors. (6) In contrast to many previous studies, our results (at larger time scales) are consistent with the expectations of economic theory: there is a positive relationship between real oil prices and real exchange rates for the major oil exporters in our dataset.

  2. THEORETICAL FRAMEWORKS AND PREVIOUS STUDIES

    2.1 Theoretical Frameworks

    In previous research, the link between the price of oil and real exchange rates has been framed by different approaches. The first centers on linking oil-price changes to balance-of-pay-ment issues and the subsequent effects on real exchange rates. The other is on the determinants of the terms of trade, which focuses on the oil price as a major determinant of real exchange rates (Benassy-Quere et al., 2007). The first theoretical approach of linking the oil price and real exchange rates to the balance of payments was by Krugman (1983) and Golub (1983). Changes in oil prices cause wealth transfers among the oil importing- and oil-exporting countries and portfolio reallocations, which results in exchange rate adjustments to clear asset markets. Krugman (1983) considers a three-country model (Europe, America, and OPEC countries) in which higher oil prices generate wealth-transfer effects from the oil importers (Europe and America) to oil exporters (OPEC). When oil prices increase, oil exporters may experience currency appreciation (since their export revenues increase), while oil importers may encounter currency depreciation through increasing energy import bills, so their current accounts would worsen. However, whether an oil-price increase causes appreciation or depreciation in an oil-importing country, or between two or more importing countries, depends on key factors, such as the relative propensity to import oil and respective bilateral trade deficits with oil-producing countries. The introduction by Golub (1983) of a fourth country (the United Kingdom) and its currency (the pound sterling, GBP) does not change the qualitative conclusions. Chen and Chen (2007) present a simple theoretical model that supports this approach. Under the assumption that the weights of the expenditure shares of traded goods are asymptotically equal, changes in real exchange rates due to oil-price changes are determined by the degree of dependence on oil imports.

    The second theoretical framework of linking oil prices and exchange rates focuses on the role of oil as a major determinant of the terms of trade (the relative price of exports in terms of...

To continue reading

Request your trial