Nonlinear and Asymmetric Impact of Oil Prices on Exchange Rates: Evidence from South Asia

1 Introduction

Oil is an important energy input for economic development of a country as oil prices affect cost of production especially in oil consuming economies. Oil price affects the macroeconomy and capital markets of the countries. Large increase in oil prices results in economic recession, surge in inflation and trade deficits in oil importing countries. It also increases the uncertainty and stocks and bonds value (Chaudhuri & Daniel, 1998). Oil price also affects exchange rate of both oil producing and consuming countries (Huang & Tseng, 2010).

Like oil, exchange rate is also important for economic growth of a country as it affects tradeable sector and investment level of the country. Fluctuations in real exchange rate (RER) occur mainly due to non-monetary/real shocks like oil prices, fiscal condition, productivity, labor supply, etc. Among these variables, oil price fluctuations is the main variable which is responsible for RER fluctuations (Chaudhuri & Daniel, 1998). Thus, both oil prices and exchange rates are associated in such a way that shocks in oil market can be transferred to exchange rate market.

Theoretically, there are different transmission mechanisms via which oil prices can affect exchange rates, i.e., terms of trade and wealth effect (Beckmann & Czudaj, 2013; Benassy-Quere, Mignon, & Penot, 2007; Habib, Bützer, & Stracca, 2016). According to the terms of trade channel, there are two sectors (tradeable and non-tradeable) in the economy and both sectors employ a tradeable input (oil) and a non-tradeable input (labor). If there is increase in oil price, then labor price should decrease so that tradeable sector may remain competitive. If non-tradeable sector is more energy intensive compared to tradeable sector, then RER will appreciate and vice versa. In oil-exporting countries, high oil prices create positive terms of trade shocks, which results in Dutch disease hypothesis by increasing the price of the non-tradeable goods which appreciates domestic currency. According to Chen and Chen (2007), in oil dependent countries, high oil price worsens terms of trade by increasing price of tradeable goods in the domestic country relatively more than in foreign country, it depreciates the RER of local currency. Further, to maintain its competitiveness, the home country needs to increase nominal exchange rate (NER), which further depreciates RER.

According to the wealth transmission channel, when oil price rises, it transfers wealth from oil-consuming to oil-producing countries. It appreciates (depreciates) the exchange rates of the oil-exporting (importing) countries by improving (deteriorating) their trade balance (Krugman, 1983). According to Golub (1983), high oil price improves (deteriorates) current account balance of oil exporting (importing) countries. If the supply of dollars in oil exporting countries is larger than the demand of dollars in oil-importing countries, then this excess supply of dollar will depreciate dollar and will appreciate non-dollar/foreign currencies. There is also an interest rate channel. According to Darby (1982), high oil price increases inflation. This high inflation increases domestic interest rate, which increases foreign capital inflows. This appreciates local currency.

After the oil crisis of 1973, many studies have investigated the impact of oil prices on different sectors of the economy like GDP, stock market, inflation, investment, exchange rate, terms of trade, interest rate, etc. Some studies also explored the influence of international oil price on exchange rates. These studies have mainly shown that high oil price appreciates (depreciates) the currencies of oil exporting (importing) economies. Section 2 elucidates literature review in detail. The existing empirical literature has mainly examined oil price and currency link for developed economies and oil-producing countries. Few authors have analyzed the case of oil-dependent small open economies. Empirical literature for South Asian countries is almost scant. Due to high economic growth, population growth, modernization, urbanization, industrialization and energy consumption, demand for oil and its related product is increasing in South Asia. The acceleration of economic growth in South Asia has increased its dependence on imported oil. The oil prices have affected the real economy of these countries by affecting different macroeconomic factors. Exchange rate is an important channel via which oil price shock is transferred to the economy. Thus, this study examines the oil price and currency linkages in oil-dependent small open economies of South Asia.

Earlier studies have used traditional econometric techniques to analyze oil price and exchange rate nexus like cointegration, Granger causality tests, the vector autoregressive (VAR) models, the vector error correction model (VECM), correlation, etc. These techniques can examine only the linear effect of oil price on exchange rate, while the actual impact of oil price on exchange rate may be nonlinear and asymmetric. Further, the traditional econometric techniques require stationarity of the variables, but both oil price and exchange rate are not stationary variables. Hence, the present study will explore the asymmetric and nonlinear influence of oil price on exchange rate using wavelet technique. This technique also does not require stationarity of the variables.

The present study bears some important contributions to the empirical literature due to several reasons. First, it will re-examine the effect of oil price on exchange rates as earlier studies have provided the inconclusive impact of oil price on exchange rates. Second, it will explore the causal relationship in time–frequency domain as causality can differ over time and frequency. For this purpose, it will use continuous wavelets tools to explore the effect of oil price on exchange rates. It will help to find nonlinear relationships among both variables. Third, the traditional linear Granger causality tests are ineffective to find nonlinear relations. Therefore, this study will use Diks and Panchenko (2006) nonlinear causality test. Fourth, it will examine oil price and currency linkages in South Asia, where only limited literature is available. Thus, this study is an important contribution to the literature.

The remaining article is laid out as follows. Section 2 elucidates literature review. Section 3 elaborates methodology. Section 4 discusses the estimated results. Conclusion is provided in Section 5.

2 Review of Literature

In literature, several studies have empirically explored the influence of international oil price on exchange rates. Earlier studies by Golub (1983) and Krugman (1983) document that high oil price may appreciate (depreciate) the currencies of oil exporting (importing) economies. Later on, several other studies have also examined oil price and exchange rate linkages. Amano and van Norden (1998) and Chaudhuri and Daniel (1998) assert that oil price fluctuation is the main factor of exchange rate fluctuation in US dollar after Bretton Woods system. Akram (2004) has shown nonlinear impact of oil price on exchange rates in Norway. Earlier literature has basically supported the Dutch disease hypothesis (Issa, Lafrance, & Murray, 2008; Korhonen & Juurikkala, 2009).

Table 1 gives the summary of the key recent studies which have explained the oil price and exchange rate linkages. The table reveals that the studies have shown mixed outcomes regarding the influence of oil prices on exchange rates. In general, the findings reveal that high oil price appreciates (depreciates) the currencies of oil exporting (importing) economies and have validated the Dutch disease hypothesis in oil exporting countries (Al-mulali & Sab, 2012; Hasanov et al., 2017; Korhonen & Juurikkala, 2009; Lizardo & Mollick, 2010; Oomes & Kalcheva, 2007). However, Mohammadi and Jahan-Parvar (2012) have provided limited support for Dutch disease hypothesis. Basher, Haug, and Sadorsky (2012) have shown that exchange rate of emerging economies become weak after oil price increase. Using data from 43 countries, Habib et al. (2016) have documented that there is no systematic association between oil price and exchange rates. Some studies have also shown insignificant association between both variables (Habib & Kalamova, 2007). According to Reboredo (2012), association between oil price and exchange rates is weak.

These inconclusive findings are due to several reasons. First, different studies have taken different sample countries which vary in their features like oil producing, oil consuming, developing, developed, emerging countries, etc. Second, different studies have used different econometric techniques for the analysis which have yielded different results. Third, studies have used different time period, data type (panel vs time series), data frequency (monthly, quarterly and annual), etc. for the analysis. Fourth, results are also different as different studies have used different (nominal vs real) exchange rates and different oil price measures.

It is apparent from the table that studies mainly concentrated on oil exporting and developed countries and ignored the analysis for small open economies. Only scarce literature is available for oil-dependent small open economies like South Asian countries. Ghosh (2011), using daily data from July 2007 to November 2008, has examined the influence of oil price shocks on Indian exchange rate. The study applied GARCH and EGARCH techniques for the analysis. The findings reveal that oil price shocks have depreciated Indian currency against US dollar. Later on, Tiwari, Dar, and Bhanja (2013a) have shown linear and nonlinear association among oil prices and Indian currency. Shahbaz, Tiwari, and Tahir (2015), using data from February 1986 to March 2009, have explored the linkages among oil prices and exchange rates in Pakistan using wavelet technique. The findings reveal that the causality among oil prices and exchange rates vary across scales. No study is available for any other South Asian country. This study will fill the gap in empirical literature by exploring the nonlinear and asymmetric impact of oil prices on exchange rates of South Asian countries by using wavelet technique. Further, nonlinear causality test of Diks and Panchenko (2006) will also be applied for robustness analysis.

Studies have mainly used traditional econometric techniques like linear regression, correlation, cointegration, Granger causality, VAR, ARDL model, GARCH, etc., to find oil price and currency linkages. Few recent studies have used sophisticated techniques like wavelet analysis, frequency domain analysis, copulas, nonlinear techniques, etc. The traditional cointegration tests are based on the assumption that the response of exchange rates to oil price shocks is symmetric. However, the response of exchange rates to oil price shocks is asymmetric. Therefore, some studies have documented that oil prices have nonlinear and asymmetric impact on exchange rates (Ahmad & Hernandez, 2013; Chen, Lee, & Goh, 2013; Kisswani et al., 2019; Tiwari et al. 2013a).

3 Theoretical Framework

To gauge the interdependence among oil price and exchange rate, both in time and frequency, wavelet technique will be used. The word wavelet means a small wave. Small means that a function is of limited length and the wave means that the function is of oscillatory type. The continuous wavelet transformation (CWT) is specified as follows (Rua & Nunes, 2009, Vacha & Barunik, 2012):

where V x ( u , c ) is estimated for the wavelet φ ( . ) for a specific time series. 1 c is the parameter to get unit variance of wavelet || φ u , c || 2 . It is a normalization factor so that the wavelet transformations may become comparable across scales and time. u ( c ) is the location (scale dilation) parameter of the wavelet. In simple words, u determines the time position and c is the scale. Scale is inversely related to frequency, where low scales explain quick changes (high frequencies), while high scales explain slow changes (low frequencies).

To measure variance at different frequency levels, wavelet power spectrum is calculated as | V x ( u , c ) | 2 . It measures the variance of series both in time and scale. To demonstrate the correlation between the series, wavelet coherence is used.

where ∁ is a smoothing operator. Smoothing is important, as without smoothing coherency would always be equal to 1. R 2 ( u , c ) falls between 0 and 1, i.e., 0 ≤ R 2 ( u , c ) ≤ 1 . Value close to zero (one) indicates weak (strong) correlation between variables in time and across frequencies.

Wavelet coherence cannot distinguish between positive and negative dependence structure as it is squared. Therefore, to get the direction of causality between the series, phase difference tool is used. The phase difference P x , y between x ( t ) and y ( t ) is represented as follows (Aguiar-Conraria, Azevedo, & Soares, 2008)

where I ( R ) is imaginary (real) part of a complex number. Two series will co-move when phase difference is zero at the specified frequency. Further, if the series is in phase (positively correlated), the arrow will be oriented to right. In this case y leads x when P x , y ∈ [ 0 , π 2 ] , and x leads y when P x , y ∈ [ − π 2 , 0 ] . The series is said to out of phase (negatively correlated), when the phase difference is π or − π . In this case, the arrows will be pointed to left. If P x , y ∈ [ − π , − π 2 ] , then y is leading, and if P x , y ∈ [ π 2 , π ] , then x is leading. Moreover, if arrows point left and up or right and down, then y leads x . If arrows point left and down or right and up, then x leads y .

Wavelet-based causality is mesaured using continuous wavelet transform without spectral matrix factorization (Olayeni, 2016), which is built on wavelet correlation method (Rua, 2013). The wavelet correlation method (Rua, 2013) is explained as

ρ xy ( u , c ) falls between −1 and 1, i.e., − 1 < ρ xy ( u , c ) < 1 . This correlation method explains co-movements of the series both in time and frequency simultaneously. The wavelet causality measure is provided as

where I x → y ( u , c ) is an indicator function such that

4 Estimated Results

4.2 Wavelet Spectra

CWT power spectral plots of oil price and exchange are shown in Figure 1. CWT plots basically show the power/variance of the series. Oil price plot indicates that it has high fluctuations between 1984 and 1986 at 0–32 months of scale (high to medium frequency), 1988–1991 at 6–14 months of scale and 2007–2009 at 8–26 months of scale frequency.^[1] Thus, oil price shows high fluctuations in three periods, but at different frequencies. In the first two periods, oil price decreased globally and in the third period, oil price surged and this period is also the period of 2007/08 financial crisis. Bangladesh exchange rate has strong power in short and medium run at 1–32 months of scale during 1983–1988. Volatility in Indian exchange rate is scattered throughout the time period at different frequency levels. Exchange rate in Pakistan has high fluctuations between 1995 and 2004 at 4–24 scales (high to medium frequency) and between 2007 and 2010 at 6–20 scales (high to medium frequency). There is volatility in Sri Lankan exchange between 1997 and 1999 at frequency level of 1–16 months.

Figure 1

CWT power spectrum of (a) oil price and (b) exchange rates. Note: Significance level at 5% is shown as thick black contour. Blue (red) color shows low (high) power. X-axis provides time period, while Y-axis provides frequency (in months).

Wavelet coherency (WTC) plots of oil prices with exchange rates are depicted in Figure 2. For WTC of oil prices with exchange rate of Bangladesh, co-movements are mainly found at frequency range of 28–36 months during 1986–1990, at scale of 16–64 months during 1993–2010, and at scale of 64 months onwards during 2002–2018. The arrows are mainly oriented to right thus variables have positive relationships. For India, co-movements are found at different frequencies from 2000 to 2010. The main co-movement is observed during 2003–2010 at 18–40 months band. For Pakistan, co-movement is found between 1995 and 2012 mainly at 8–16 months frequency and at 2–32 months frequency. As for as Sri Lanka is concerned, there are co-movements between oil price and exchange rate at different frequency levels at different time periods but the main co-movement between both the series is found at 36–68 months frequency during 1990–2010 time periods. The arrows point to right and up, it shows that oil price has positive influence on exchange rate.

Figure 2

WTC between oil prices and exchange rates. Note: Significance level at 5% is shown by thick black contour which is calculated using Monte Carlo simulations.

Wavelet causality results are provided in Figure 3. The color code indicates the strength of causality and it goes from 0 to 1. For Bangladesh, a strong causality among oil price and exchange rate is observed between 2003 and 2010 on 20–32 months scale and between 1998 and 2008 on 64 months onward frequency. This causality is little bit weak in India and Sri Lanka and is strong in Pakistan where a strong causality holds between 2006 and 2011 at 4–16 months. Causality also holds at different time periods and at different frequency levels. Our results support the findings of Tiwari et al. (2013a), which also did not find any causality between oil price and Indian currency at low time scales.

Figure 3

Wavelet-based causality from oil price to exchange rate. Note: Statistical significance at 5% (10%) is shown through white (red) contours.

Figure 4 reports the Rua (2013) measure of CWT correlation. It is evident from these figures that variables are positively correlated during the same period as shown in WTC plots given in Figure 2. These results generally support the findings of WTC. The first figure gives the correlation among oil price and exchange rate in Bangladesh. Both variables are positively correlated during 1995–2001 at 8–24 months scale (high and medium frequencies) and 2002–2010 at 0–32 months band. This positive co-movement also holds for most of the time period at low frequencies. In India, the correlation between series is observed at high to medium frequency 0–24 months scale but at different time periods. The main strong correlation is found during 2006–2010 at 5–16 months scale. For Pakistan, correlation is found between 1993 and 2003 at 6–24 months scale and between 2006 and 2011 at 2–16 months scale. Correlation is also found at low frequency from 1983 to 2004. In Sri Lanka, correlation among oil price and exchange rate is found at different frequency levels and at different time periods. Correlation holds during 1989–1993 at 4–16 months frequency level and during 2004–2010 at 2–16 months scale. A very strong correlation between both series is found during 1993–2018 at 32–100 months frequency (low frequency). It indicates that oil price has affected South Asian currencies. These correlation results support the notion that oil price fluctuations are positively associated with exchange rates in South Asia.

Figure 4

Wavelet-based correlations (Rua, 2013) between oil price and exchange rates. Note: The color code shows strength of correlation. Blue (read) color indicates negative (positive) correlation between variables.

5 Conclusion

The study examines the association between oil price and exchange rates of South Asian economies using data from July 1983 to June 2018. For empirical analysis, wavelet coherence framework is used as it provides relationship among series in time–frequency space. The findings indicate that oil price and exchange rates are positively correlated. It shows that high oil price depreciates South Asian currencies. Being oil importing economies, this result is in line with theoretical expectations. The results of nonlinear causality show that there is bidirectional causality among oil price and exchange rate in Bangladesh and India and unidirectional causality between oil price and exchange rate in Pakistan and Sri Lanka.

These results have some useful policy recommendations for investors and policy makers. Monetary authorities in South Asian countries need to consider the role of oil price shocks in determining exchange rate dynamics. To stabilize the exchange rates after oil price shocks, monetary authorities should formulate and implement policies in such a way that the impact of oil price shocks on exchange rates is minimized. Monetary authorities may concentrate on inflation target to offset exchange rate instability. Further, South Asia countries should increase domestic production of oil and look for other energy sources to prevent depreciation of their currencies. High oil price can misalign the exchange rates. The flexible exchange rate system of South Asian countries can make their economies susceptible to changes in oil price shocks, as in flexible exchange rates system high oil price produces greater volatility in exchange rate compared to fixed exchange rate system. Therefore, these countries may use managed flexible exchange rate system to avoid exchange rate volatility after oil price shocks.