The oil price and exchange rate relationship revisited: a time-varying VAR parameter approach.

• Author: Bremond, Vincent
• Position: Vector auto-regressive - Report

1. Introduction

   1.1 Context

   Since the beginning of the 1970s, foreign exchange and oil markets both sustained shocks and crises. With the unilateral cancellation of the direct convertibility of the US Dollar to gold on August 15, 1971, the world economy experienced a new financial context with the advent of a floating foreign exchange market. In 1976 the Jamaica Agreement replaced defacto the Bretton Woods System based on exchange rate stability. In the oil markets, the progressive take-over by the Organization of the Petroleum Exporting Countries (OPEC), at the beginning of the 1970s, and the first oil shock in 1973-1974 can be considered as the starting point of a new context of oil price volatility compared to the previous decade. As oil is quoted in US dollars (henceforth, USD), the oil prices and the value of the dollar seem to have a mutual influence and it is relevant to study the interaction between those two variables. Several studies have previously investigated the link between the USD and the oil prices.

   1.2 Literature Review

   Krugman (1980, 1983) analyses the variation of the USD following an oil price increase in a model with three different areas (America, Germany and the OPECmembers). He demonstrates that the final effect on the Dollar exchange rate in the short run heavily depends on the comparison between the US weight in world oil imports and the share of Dollar assets in OPEC's portfolio; while in the long run, the comparison has to be made between the US share in world oil imports and the weight of US goods in OPEC's imports. Golub (1983) divides the world in three areas (America, Europe and the OPEC countries) and two currencies (USD and Deutsche Mark) and focuses on the wealth transfers to the OPEC countries. Under the assumption of inelastic demand of oil from Europe and America, the USD depreciates against the German currency if the OPEC countries have a higher propensity to hold Deutsche Marks than the oil-importing countries. If, as a result of the wealth transfer, there is an excess demand for Deutsche Mark, then the dollar exchange rate will depreciate.

   There is no clear consensus in the literature concerning the direction (influence of the dollar on the oil price or vice versa) and the sign of the relationship between oil price and the value of the Dollar (e.g., Beckmann and Czudaj, 2013). Regarding the direction of the relationship, some studies focused on the oil price to exchange rate causality. Chen and Chen (2007) investigate the long-term relationship between the real oil prices and the real exchange rate. Using a panel data methodology, they find a cointegration relationship using a set of different markers of crude oil or basket price (Brent, Dubai, West Texas Intermediate (WTI) and world prices), and conclude that the oil prices are "the dominant source of real exchange rate movements". However, Sadorsky (2000) and Krichene (2007) among others find a reverse causality from USD to oil prices. Therefore, the direction of the causality between oil price and exchange rate is not clear cut. Percebois (2009) summarizes the complexity of the relationship by emphasizing the possibility of a bilateral causality through various macroeconomic channels such as: the effect on world demand through local prices, a Chinese effect, a petrodollar recycling effect, and a target revenue effect. Benassy-Quere et al. (2007) highlight the existence of a cointegration relationship between the oil price and the Dollar real effective exchange rate for the 1974-2004 period with the causality link running from oil price to the exchange rate. Nevertheless, the authors suggest that the causality link could be reversed in the 2002-2004 period because of the Chinese monetary policy. Benhmad (2012), using wavelet analysis, also emphasized that time scales matter. Causality between real oil price and real effective US Dollar exchange rate returns runs in bidirectional ways in long time horizons, while in short time horizons, causality runs from oil prices to real effective US Dollar exchange rate returns.

   Regarding the sign of the relationship, there is also no clear consensus in the empirical literature. Amano and Van Norden (1998a, 1998b) show that the impact of an oil price increase on the exchange rate is heterogeneous across countries: a 10% increase in the oil price leads to a depreciation of both Japanese and German currencies (respectively of 1.7 and 0.9%), when it causes an appreciation of the USD of around 2.4%. Relying on two monetary models (Basic and Composite Models), Lizardo and Mollick (2010) highlight that the dependency to oil could play a major role in the relationship. They split the country panel into two groups (crude oil net importers or net exporters). They demonstrate that an increase in the oil price is followed by an appreciation of the net exporters' currencies relative to the USD, while for the oil importing countries, their currencies tend to decrease relative to the USD. Reboredo et al. (2014) study the relationship between WTI price and the exchange rate of US Dollar against a set of currencies (1), splitting the period of studies into two samples: one before the crisis (July 2008) and one after. The authors find that the cross-correlations between oil price and exchange rates are negative and that the dependence increases after the start of the financial crisis. With an analysis based on impulse responses and forecast error variance decomposition, Akram (2009) focuses on the relationship between oil prices, commodity prices, the USD exchange rate, the global output and interest rates. Building three VAR models, he argues that a positive shock on the oil price is followed by real exchange rate depreciation, while a real exchange rate depreciation leads to higher commodity prices (including oil prices). Creti et al. (2013) using a dynamic conditional correlation (DCC) GARCH methodology investigate the links between 25 commodities and stocks from 2001 to 2011 and conclude that the correlations among those variables are not constant over time and are highly volatile since the financial crisis. Finally Jebabli et al. (2014) using a new time varying parameter VAR (TVP-VAR) model with a stochastic volatility approach for a large set of commodities (crops, livestock, plantation and forestry products) analyse the spill-over effect between financial, food and energy markets. They demonstrate an increase of markets contagion since 2008.

   Thus, there is no clear-cut result regarding neither the direction of the relationship between exchange rate and oil price, nor the extent of the elasticity due essentially to different sample period considered in the previous listed studies. In our paper, we analyse the oil-dollar link with an innovative methodology. To improve the comprehension of the oil price-exchange rate relationship, we use time-varying Bayesian VAR to estimate the different parameters of interest. This methodology offers four main advantages in our analysis. First, as a VAR analysis, it takes into account endogeneity among the set of variables and, thus, permits to circumvent the absence of consensus concerning the direction of the relationship. Second, the estimated elasticities, which measure the movement in percentage of the variables of interest following one percent change of one variable, evolve through the whole estimation period, enlarging the scope of the interpretation (in terms of magnitude and periodicity) compared to previous empirical studies. Third, it permits to model both abrupt break and persistence in the relationship between variables of interest. Indeed, relationship between oil price and exchange rate might be subject to structural break due to different exogenous oil events such as geopolitical events, OPEC decisions, economic or monetary policy changes, etc. It might also be subject to gradual evolution due to adaptative learning behaviour of agents (Primiceri, 2005). Therefore, letting data determine whether the relationship presents a break or is persistent is a valuable future of this methodology. Fourth, the Forecast Error Variance Decomposition (henceforth FEDV) which measures the contribution of different variables in the model to the volatility of the variable of interest is calculated for the entire period of the sample. Up to our knowledge, this paper is the first one that uses the time-varying FEDV for the oil pricesexchange rate analysis and, hence, contributes to a more comprehensive analysis of this subject with regards to time-changing economic environment pattern and monetary policy during the estimation period.

   We use monthly data on oil price (Spot market Brent price expressed in US dollars) (2), the gold price (spot price in the London Commodity Exchange measured in US dollars), the effective exchange rate of the Dollar, the HWWI index (3) (Industrial raw materials measured in US dollars) and the dry cargo index as a proxy for the real economic activity. Our monthly data cover the sample period from 1976:07 to 2013:07. Oil and gold prices are extracted from Datastream database (respectively, with code "UKI..C..A" and "UKOILBREN"). The US Dollar effective exchange rate we use is the nominal major currencies Dollar index extracted from the board of governors of the Federal Reserve System (Foreign Exchange Rates - H.10). Finally, the Dry Cargo index is available at Lutz Kilian personal webpage. Except for Dry cargo index, all variables are rendered stationary by taking the first difference of their natural logarithm. The rest of the paper is organized as follows: Section 2 advances arguments why the link between oil prices and macroeconomic and financial variables, more particularly the exchange rate, is time-varying; Section 3 is dedicated to the model specification; Empirical results are provided in Section 4; Section 5 concludes the article.

2. Why the link between oil prices and exchange rate has evolved over time?

   In...