**1. Introduction**

Over the last decades, the oil market registered significant growth, becoming the world's biggest commodity market and transforming from a purely physical to a highly sophisticated and complex financial market [1]. Its rhythm of growth remains high: the global oil and gas market is expected to grow from \$4677.45 billion in 2020 to \$5870.13 billion in 2021 at a compound annual growth rate (CAGR) of 25.5%, and the market is expected to reach \$7425.02 billion in 2025 at a CAGR of 6% [2]. In addition, crude oil also tops the commodities markets in terms of liquidity, being the most actively traded commodity around the globe, while the price of oil reflects the overall health of the energy sector worldwide.

Oil price forecasts are of crucial importance for macroeconomic projections, which is especially explained by the impact that oil prices have on inflation and output and,

**Citation:** Tudor, C.; Anghel, A. The Financialization of Crude Oil Markets and Its Impact on Market Efficiency: Evidence from the Predictive Ability and Performance of Technical Trading Strategies. *Energies* **2021**, *14*, 4485. https://doi.org/10.3390/en14154485

Academic Editor: Tomasz Korol

Received: 22 June 2021 Accepted: 21 July 2021 Published: 24 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

consequently, on the issuance of monetary policy. However, recent movement of crude oil markets has highlighted the difficulty in forecasting oil prices and attested that oil market dynamics tend to vary substantially over time. Moreover, crude oil markets are characterized by increased volatility, which might be explained both by variations in the price elasticity of oil demand and supply, and also by the process of "financialization" of the oil market with the increasing use of oil as a financial asset [3–6]. Consequently, oil derivatives markets have expanded over the last decades, with the presence of purely financial practitioners (institutional investors such as hedge funds, pension funds, insurance companies, and also individual traders) with no interest in the physical commodity becoming more prominent. Concurrently, a variety of instruments that permit speculation in oil have become available for trading, from passive investment vehicles such as energy indexes and ETFs to derivative instruments such as futures, options, or CDFs. All these developments in oil markets have a direct impact on the oil market movements, its efficiency and subsequent predictability.

Financial institutions and regulators around the globe (i.e., The Federal Reserve Board, the World Bank, the International Energy Agency, the European Central Bank etc.) regularly issue oil price forecasts, which is further a paramount factor for policy formulation within the European Central Bank (ECB), the IMF and the Federal Reserve Board [7]. However, predicting oil price movements remains a challenging endeavor [3], which is further complicated by its increasing financialization and the intense speculative activity within the market that improved its efficiency (in Fama's EMH sense) and hence contributed to its unpredictability. Moreover, none of the techniques previously employed for oil price forecasting has proved particularly successful and thus presently there is no "optimal" or commonly accepted forecasting technique for oil price [8].

As such, the analysis of the efficiency of the crude oil markets is a timely research topic, with important implications for policy issuers and for financial markets practitioners. Nonetheless, and somewhat surprisingly given the practitioners' interest in this commodity as reflected in its market liquidity, the academic literature on the profitability of technical trading rules applied to crude oil markets remains rather scarce. Our study contributes to extending this literature. This paper thus revisits the Fama efficiency [9] of the crude oil markets though exploring the predictive ability and trading performance of a plethora of technical trading rules (TTRs) applied to relevant energy series (i.e., WTI crude, Brent crude and XTE). Moreover, our focus on energy/oil markets is even more motivated by the fact the COVID-19 pandemic has severely impacted the oil markets, due to travel restrictions, disrupted supply chains and imposed government lockdowns. Previous studies have found that the efficiency of crude oil markets is lost during crisis periods, investigating the 2008 global financial crisis [10,11]. As the impact of the ongoing pandemic crisis on the oil market efficiency has not been yet assessed, this constitutes a secondary research goal of the current study and a further contribution to the extant literature.

The Efficient Market Hypothesis (EMH) and the related concept of market efficiency remain paramount in modern finance, with a plethora of empirical studies dedicated to confirm it on different markets, assets and time periods, with divergent results. EMH has its roots in the works of Eugene Fama [12,13] and Paul Samuelson [14]. Furthermore, Fama's seminal work defines an efficient market as "a market with a large number of rational, profit "maximisers" actively competing, each attempting to predict future market values of individual securities, and where current important information is almost freely available to all participants" [15] and it also distinguishes between three forms (or 'strengths') of market efficiency—weak, semi-strong and strong. In its weak-form, EMH states that current prices reflect all existing historical information, and thus prices will exhibit random walk.

Alternatively, technical analysis (or Chartism) specifically involves making investment decisions based on past price movements. As Alexander [16] has said it, "the technician studies price movements of the immediate past for telltale indications of the movements of the immediate future." However, in relation to EMH, this would imply that technical trading rules (TTRs) based on historical price data would offer no predictive power, and hence technical analysis would be inexpedient. Nonetheless, as Menkhoff [17] shows, technical analysis remains very popular among practitioners, with the vast majority of 692 surveyed fund managers from five countries acknowledging relying on technical analysis for market timing and decision-making, and to favor it relative to fundamental analysis. This is an indication that Chartism must hold some value to traders that is unaccounted by the EMH.

Consequently, in this study, we choose to employ instruments pertaining to technical analysis (i.e., TTRs) to investigate the overall efficiency of the oil markets, to assess the potential differing financialization process of the two most important crude oil markets (WTI and Brent) and to analyze the impact of the COVID-19 induced crisis on oil markets' efficiency.

The remainder of the paper is organized as follows. The next section gives a review of the literature concerned with technical trading rules applied to commodity markets and most specifically their predictive ability and performance on oil/energy markets. Section 3 discusses the data and method. Empirical results and discussions are contained in Section 4, followed by some concluding comments in the final section.

#### **2. Literature Review**

Although the predictive ability and profitability of technical trading rules applied to various international stock markets during different time periods have been thoroughly examined, the literature on technical trading rules applied to commodities markets in general and energy/oil markets in particular remains rather scarce.

One of the first studies in this narrow literature is that of Marshall et al. [18], which test over 7000 rules on 15 commodity futures markets, including WTI crude oil, heating oil and soybean oil for a period spanning 1 January 1984–31 December 2005. They analyze the entire series and two equal sub-periods and cannot report that technical rules achieve superior performance after accounting for data snooping, except the oats market. The oil markets are thus found to be efficient over the 1984–2005 period.

Further, Szakmary et al. [19] examine the profitability (net of transactions costs) arising from the implementation of 12 trading rules (six DMAC and six channel specifications) on a monthly dataset for 28 commodities, having a different start date for each series and with all series ending on 31 December 2007. The dataset includes the same three oil markets again, i.e., WTI crude oil, heating oil and soybean oil, and results confirm that technical rules do perform well, although mean returns are lower and less significant toward the end of the analyzed period (i.e., during the 1996–2007 sub-period), especially when testing is restricted to high-volume markets, a category to which WTI crude oil belongs. Nonetheless, the authors refrain to claim that their study confirms the weak-form inefficiency of commodity futures markets included in the analysis.

Narayan et al. [20] use daily data on four commodities, including again WTI crude oil, spanning the period 16 May 1983–22 November 2011, to which they apply a narrow universe of six standard moving average (SMA) trading rules and report that investors can earn abnormal return (net of commissions) from technical trading rules in three of the four markets, including in the WTI crude oil market, where trading rules achieve the highest return. However, their results do not seem to account for data snooping, which is a bias proven to have a significant impact on results and thus are not sufficiently strong to prove the inefficiency of the WTI crude oil market. Subsequently, Narayan et al. [21] also conclude that commodity futures markets can indeed offer investors statistically significant profits.

Further, Wang et al. [22] employ daily prices of WTI crude oil futures contracts over 1983–2014 and develop dynamic MA trading strategies through genetic algorithms, whose trading performance is further compared to the buy-and-hold strategy and to some static MA rules. The study confirms the superiority of dynamic moving averages on the WTI crude oil futures market during downward trending markets. However, it also lacks a check of results robustness.

More recently, Psaradellis et al. [23] offer probably the most updated study on technical trading rules applied to the crude oil market. The study thus investigates the success of the 7846 trading rules proposed by Sullivan et al. [24] applied on the daily prices of WTI crude oil futures and on the United States Oil (USO) fund, from 2006 to 2019. Results confirm that there is no persistent nature in rules' performance for the two oil markets after adjusting for data snooping, thus supporting WTI market efficiency for the 2006–2019 period, although some interim market inefficiencies might be encountered.

Overall, previous studies thus generally agree on the efficiency of the WTI crude oil market for different periods, all spanning before the ongoing COVID-19 pandemic, after adjustment for data snooping-bias is made. To the best of our knowledge, the efficiency of the Brent crude oil market in relation to the performance of technical trading strategies has not been tested, nor has the efficiency of the XLE fund market. This study intends to fill this void, providing relevant results for policy makers, academics and investment practitioners.

Thus, we add to the literature first by extending the energy markets under scrutiny by including the most traded crude oil contract at world level, i.e., Brent crude oil along with a relevant energy-traded ETF, namely XLE and, secondly and most importantly, by an updated investigation on the performance of a large universe of TTRs during an historically turbulent period for crude oil markets and energy portfolios (i.e., the COVID-19 pandemic).

Additionally, a non-trivial issue about TTRs and their performance that needs further discussion is testing the statistical significance of results.

In this respect, bootstrapping firstly emerged as a convenient way of testing TTRs on data generated using some algorithm. Brock, Lakonishok and LeBaron [25] proposed the bootstrapping methodology for testing the predictability of some of the simplest trading rules and found that technical rules—in particular SMA—were able to achieve excess returns that could not be explained by a random walk model, an AR (1) process, nor a GARCH (M or Exponential) model. Another method, the stationary bootstrap that resamples from blocks of data with random lengths, was developed by Politis and Romano [26]. However, the bootstrapping methodology developed by Brock et al. [25] is the one that has been extensively applied in the literature concerned with the profitability and predictability of TTRs on speculative markets.

Nonetheless, this method is vulnerable to the so-called data-snooping bias. Data snooping reflects the process of testing and retesting filters, rules and combinations on a high number of randomly generated series until some (apparently) significant specifications emerge. In other words, the data snooping bias reflects the danger that the best forecasting model encountered in a specification search is just the result of chance instead of superior forecasting abilities and thus has no predictive superiority over a given benchmark model. Among others, Fang et al. [27] demonstrate that the predictive ability of the technical trading rules employed by Brock et al. [25] disappears when the sample selection bias, data mining, hindsight bias, and other usual biases are accounted for. Park and Irwin [28] also confirm that most studies that do encounter superior profitability of TTRs are subject to various problems in their testing procedures, including biases, which should be addressed in order to provide conclusive evidence. In addition, Harvey and Lu [29] draw attention that seemingly successful trading strategies can be encountered by chance, and the "no– biases" assumption of traditional tools of statistical analysis no longer hold.

The first strong solution for the data-snooping danger, still seen as the standard method for adjustment, was proposed by White [30], and was based on results from Sullivan, Timmermann, and White [24]. The procedure, entitled White's Reality Check (RC) for data snooping, tests the null hypothesis that the best model does not have predictive superiority over a benchmark versus the alternative that the best model is over performing.

Afterwards, there have been some attempts in the literature to improve this methodology. Mainly, Hansen [31] maintains that the RC procedure can be affected by testing a large plethora of irrelevant rules, an issue that can be corrected by the "Superior Predictive Ability" (SPA) test. Further Bajgrowicz and Scaillet [32] introduce the false discovery rate

(FDR) as a new approach to data snooping and show that the economic value of TTRs that has been previously reported in the literature is no longer significant.

In this paper, we proceed to check the robustness of our results first by applying the popular Brock et al. [25] bootstrapping methodology (on a higher number of randomly generated series than employed by the original study and most others thereafter) and further, we correct for the data-snooping bias by following the most commonly used technique, namely, the RC procedure. This approach has the advantage of allowing easy comparison of results with previous related studies, and thus contributes to a higher relevancy of results.

#### **3. Data and Method**

#### *3.1. Data*

In the empirical modeling, we use daily spot prices of the two main grades of crude oil (Brent and WTI), as well as daily prices of a representative energy-traded fund, the Energy Select Sector SPDR® Fund or XLE. As XLE has the smallest trading history, to ensure comparability across markets, we set the same data window for the three time series, and hence data will span 1 January 1999 through 29 March 2021, or a total of 5686 daily observations for each energy market.

Brent North Sea Crude (also known as Brent crude oil) and West Texas Intermediate (known as WTI crude oil) are the most widely traded oil grades. Brent Crude is produced in the North Sea between Shetland Islands and Norway, while West Texas Intermediate is produced in the United States fields. According to the US Energy Information Administration, "sweet crude" refers to crude oil that has sulfur content of less than 1%, a category that Brent and WTI both fall under. Furthermore, both are less thick (or "lighter") than other types of crude oils, making them quicker to process and thus more appealing to manufacturers of petroleum products. Brent crude is the reference price for crude oil in Africa, Europe, and the Middle East, and it is assumed that Brent determines the value of around two-thirds of global crude oil production. Alternatively, West Texas Intermediate stands as the major oil benchmark for North America. As far as trading crude oil is concerned, Brent crude oil is listed on the New York Mercantile Exchange (NYMEX), a division of the Chicago Mercantile Exchange (CME), whereas Brent is listed on the electronic Intercontinental Exchange (ICE). As a result of their respective host markets, delivery locations vary by country in the case of Brent crude, which is traded internationally, while the main delivery location for physical exchange and price settlement for WTI is Cushing, Oklahoma. The price differential between Brent and WTI (which is a consequence, among others, of different transportation costs, of the supply and demand balance in different parts of the world, of geopolitical events, etc.) is called a spread.

The Energy Select Sector SPDR® Fund (XLE) mirrors the S&P 500's market-capweighted index of US energy companies. The Select Sector SPDR Exchange Traded Funds divide the S&P500 into nine industry categories, with XLE representing the energy sector. As a result, XLE is an investment vehicle that provides traders with a desired exposure to firms in the oil, gas, and consumable fuel industries, and related services.

Crude oil prices are obtained from the Federal Reserve Bank of St. Louis's (FRED) database, which collects data from the U.S. Energy Information Administration, while data for XLE are collected from Yahoo! Finance.

We argue that a separate investigation for a recent and relevant time period (the 2020–2021 COVID-19 pandemic) is not only more appropriate, but also more relevant to academics and investment practitioners. We base our hypothesis on previous empirical findings on the performance of TTRs on energy markets that show that the returns to technical strategies are not consistently strong for periods up to 2005 [18] or up to 2019 [23]. Thus, in order to take a closer look at the turbulent ongoing pandemic period, we will subset the so-called "COVID-19 window," which is spanning 1 January 2020 through 29 March 2021.

Figure 1 reflects the evolution of the BRENT and WTI crude oil price from January 2020 to March 2021, showing historical lows and significant volatility during the pandemic period. On 20 April 2020, the WTI crude oil price was disconnected from its typical relationship with the price of Brent crude oil, collapsing by more than \$50/barrel.

**Figure 1.** Spot Prices (Crude Oil in Dollars per Barrel) during the pandemic period (January 2020– March 2021). Source of data: U.S. Energy Information Administration, Crude Oil Prices: West Texas Intermediate (WTI)—Cushing, Oklahoma and Europe Brent, retrieved from FRED, Federal Reserve Bank of St. Louis; Author's representation.

Overall, the price of both WTI and Brent crude oil during the pandemic period registered a dramatic fall in the early stages of COVID-19 up until April 2020 and a subsequent recovery to pre-pandemic levels by March 2021, attesting the efficiency of interventions by oil-producing countries that have imposed supply caps, and also reflecting the optimism about post-pandemic economic recovery resulting from the progress of COVID-19 vaccine distribution worldwide.

The three daily energy series are turned into daily returns indexed from R to T, so that T=R+n − 1. We follow White [27] and compute daily returns as:

$$y\_{i,t+1} = \frac{Index\_{i,t+1}}{Index\_{i,t}} - 1 \tag{1}$$

where *yi*,*t*+<sup>1</sup> is the return of the Index *i* on trading day *t* + 1.

Figure 2 provides an overview of the three energy markets average return volatility over the 1999–2021 period, attesting the particularly dramatic month of April 2020, especially in the case of the WTI crude oil market.

The summary of descriptive statistics for one-day buy-and-hold returns for all three energy series employed in the empirical estimations, for the pre-pandemic period and also for the COVID-19 window, are presented in Table 1, panel A and panel B, respectively.

**Figure 2.** WTI, Brent, and XLE Average Monthly Returns (January 1999–March 2021). Source of data: Author's representation with crude oils daily price data sourced from the U.S. Energy Information Administration, retrieved from FRED, Federal Reserve Bank of St. Louis and XLE daily price data sourced from Yahoo! Finance.


**Table 1.** Descriptive statistics for one-day returns for WTI, Brent, and XLE.

The mean daily returns for the energy series largely confirm common perceptions of these markets. The XLE fund shows returns that compare rather well with the crude oil series during the whole 22-year period, and it also presents the lowest volatility of price returns both before and during the COVID-19 pandemic. On the other hand, during the pandemic period, WTI is the least rewarding in terms of return and also the riskier in terms of volatility among the three series. The Brent crude oil market has the highest mean returns for the pre-pandemic period (of about 0.07% per day) and also for the COVID-19 window (0.13% per day), while WTI is the only market that lost in terms of daily returns over the 2020–2021 period, whilst also being the most risky market. WTI statistics are surely strongly influenced by the historical plummet that the WTI price has suffered in April 2020. We notice from data presented in Panel B of Table 1 the dramatic aforementioned daily

drop of over 300% for WTI crude oil prices in April 2020, the largest one-day decrease in history.

This is further also more clearly reflected in Figure 3, showing returns volatility for the three energy markets during the pandemic period.
