Macroeconomics
EvolvingOil-DollarDynamicsAmidHeightenedMacroUncertainty
5 mins
As a barometer for the latest macro developments, the swings in crude oil prices underscore the lingering uncertainty posed by the pandemic. In July, West Texas Intermediate crude eclipsed $75 a barrel, marking the highest price since October 2018. It subsequently tumbled amid surplus supply and the spreading Delta variant, which renewed global concerns about lockdowns and travel restrictions. Given the recent volatility, this post examines the well-known relationship between the movements of oil and the U.S. dollar, with the goal of determining whether the oil price shock has upended the dynamics between the two assets.
At a high level, we view the oil-dollar relationship through the lens of exchange-rate pass-throughs: Given the U.S. dollar is often the currency of choice in global trade, how do fluctuations in dollar exchange rates impact the local price of goods? In the case of oil, do importers simply accept higher prices when converted to their local currencies, implying a minimal correlation between crude and the dollar? Or do oil exporters accept lower dollar prices to shoulder some of local consumers’ pain, a scenario that denotes a negative correlation between the two?
As a first pass, we look at how the correlation between oil prices and the DXY Dollar Spot Index has evolved over the past decade (Figure 1). A few observations stand out. First, we note a relatively strong negative correlation between the two series, although the strength of the relationship varies throughout the sample period. Second, this negative relationship became much weaker in the immediate aftermath of the COVID pandemic and oil price shock, when prices fell precipitously and even briefly went negative.1
Figure 1
WTI Oil vs. DXY Dollar Spot Index and Their Correlation
Bloomberg, Federal Reserve, Haver Analytics, and PGIM Fixed Income. Grey bar denotes the pandemic period.
With this starting point in mind, we further investigate this relationship using a simple regression framework. Our approach is anchored by former Fed Chairman Ben Bernanke’s research, which was modeled off that of macroeconomist James Hamilton, who proposed an empirical basis for thinking about these dynamics.2 We model changes in the price of oil using changes in the USD spot index, copper prices, and the 10-year Treasury yield. In a modified model, consistent with Bernanke’s work, we also include changes in the VIX Index to account for market risk sentiment. These variables provide a first step toward disentangling the dollar’s impact on oil prices from other factors, allowing us to get a cleaner reading on the dynamics of interest.
Bernanke’s sample runs from mid-2011 to mid-2014 and uses daily data. The coefficients on all variables are statistically significant and have the expected signs. In his baseline model, the coefficient on the dollar variable suggests a 1% increase in the dollar is associated with a 0.74% decrease in the price of oil, holding other variables constant.
We update Bernanke’s results to examine whether this statistical relationship has changed. For consistency, we also start the sample in mid-2011. However, given the distortions observed in Figure 1 associated with the pandemic and oil price shock, we analyze how the coefficient of the dollar has evolved since the onset of these developments. Accordingly, we proceed by using two estimation periods: one ending in February 2020 and a second ending in July 2021.
Figure 2 displays the results, which are broadly consistent with those in Bernanke’s sample. Specifically, in the truncated sample, the dollar continues to have a statistically significant and negative coefficient, although the magnitude is slightly lower than that observed in the earlier sample period. Still, our baseline model suggests that a 1% appreciation in the dollar is associated with a 0.61% decrease in the price of oil, holding other variables constant. Considering that changes in exchange rates do not entirely pass through to oil prices, it appears that oil exporters and importers share the burden of exchange-rate fluctuations. That is, when the dollar appreciates, exporters generally accept a lower dollar-denominated price while importers generally appear willing to pay more in local currencies.
Figure 2
Model Results
Truncated sample is July 2011 to February 2020, full sample is July 2011 to July 2021. Model estimated with Newey-West standard errors; t-statistics are in parentheses; * denotes significance at the 1% level.
However, when extending the analysis through July 2021, we observe that, while the dollar variable still has the expected sign, it becomes statistically insignificant in both models. This motivates us, as a further check, to acknowledge a possible breakdown in the negative relationship during the early months of the pandemic.
Accordingly, we run a follow-up exercise using rolling regression analysis. A backward-looking window of 200 days was adopted, in which the coefficients are re-estimated within each iteration. Figure 3 plots the rolling coefficient estimates of the dollar variable. Prior to March 2020, the interquartile range of the coefficient was negative as expected, ranging from -0.07 to -0.77. However, the oil price shock and pandemic threw this relationship in flux, driving the estimated coefficient well above what historical patterns would suggest. One explanation is that variables tend to experience drastic moves during unexpected exogenous shocks and historical correlations often break down. Indeed, Figure 1 highlights a brief spike in the oil-USD correlation once mid-March 2020 entered the rolling window. A related explanation is that a linear model is not well-suited to explain the macroeconomic dynamics witnessed last year. Nevertheless, we find that a reversion has occurred more recently, suggesting a return to the more-familiar inverse relationship as some of the shorter-term economic difficulties subside.
Figure 3
Time-Varying Coefficient of DXY
Bloomberg, Federal Reserve, Haver Analytics, PGIM Fixed Income. Grey bar denotes the pandemic period.
Overall, it is informative to view these results in the context of exchange-rate pass-throughs. The regression evidence suggests that oil producers and consumers share the impact of exchange rate volatilities, which strikes us as conceptually plausible and broadly consistent with observations of exchange-rate pass-throughs in other settings. This coefficient isn’t equal to one (full pass-through of dollar moves) because the higher oil price tends to edge down demand and reduce the real incomes of importers, incentivizing them to seek substitutes and economize their existing resources. Conversely, the coefficient isn’t closer to zero because oil exporters have market power and are able to pass on some degrees of price increases. Finally, the time-variation observed in the estimates could reflect varying business cycle and oil market conditions, as well as normal variations in economic behavior and the global backdrop more broadly.
The oil-dollar relationship remains a key factor for crude exporters and importers throughout the global economy. While the expected negative correlation exists over prolonged periods, it deteriorates when certain shocks are introduced. Hence, as we monitor the credit conditions across oil-sensitive economies, we will continue to evaluate how the pandemic and other potential shocks might affect this key relationship.
1 Dezember, Ryan. U.S. Oil Costs Less Than Zero After a Sharp Monday Selloff. Wall Street Journal. April 21, 2020.
2 Bernanke, Ben. The relationship between stocks and oil prices. Brookings Institution. February 19, 2016.