Abstract

The influx of institutional investors in commodity markets has increased the complexity faced by economic agents and policymakers. Increasing uncertainty that originates from economic or non-economic uncertainty shocks contributes to increasing food price volatility and market speculation. Acting as speculators, institutional investors with large volume trading led commodity prices and volatility to higher levels in 2007-2008. In this paper we argue that economic agents and policymakers face difficult challenges aiming to address those problems related to uncertainty. The lack of probability distributions that can characterize uncertainty prevents economists from developing compatible predictive models. Modeling uncertain economic phenomena with the use of the sentiment and behavioral aspects of economic agents has not proved sufficient to improve the prediction ability of economic models. Thus, we need novel approaches to analyze commodity markets and model choice behavior.

Keywords:Uncertainty; Commodity financialization; Convenience yield; Futures prices; Sentiment

Abbreviations:CITs: Commodity Index Traders; CY: Convenience Yield; ARDLX: Autoregressive Distributed Lag with Exogenous Variables; VIX: Volatility Index; TRMI: Thomson Reuters Market-Psych Indices.

Introduction

Portfolio diversification is a simple strategy whereby investors spread their money over a number of different assets to reduce and manage overall investment risk. Such assets include individual securities in broad categories of stocks, bonds and Treasury bills. In the last two decades of the 20th Century, commodities (grains, energy products, metals) have entered the investors’ radar as they were offering opportunities for additional diversification benefits. Indeed, commodities exhibited unique risk and return characteristics different from the characteristics of stocks, bonds and bills. As a result, commodities became another distinct investment category. Since early 2000s, portfolio and institutional investors start taking positions in the commodity futures markets [1] while we observe extensive use of oil derivatives as a financial asset by hedge and pension funds, insurance companies, retail investors and other market participants [2]. Furthermore, Commodity Index Traders (CITs) invest in a broad portfolio of commodities constructed as an index that contains energy, metal and agricultural commodities [3]. The inclusion of commodities in portfolios along with financial assets gave the term “commodity financialization.”

Since financial investors lack physical positions in commodities, their positions in futures markets are considered speculative. Also, because financial investors trade in large volumes, their presence in commodity markets adds to the complexity of financial investors’ decision making under risk and uncertainty. Moreover, their behavior directly affects commodity futures and cash prices as well as commodity inventories. Additionally, it indirectly affects commodity price volatility which is a significant source of uncertainty for economic agents such as investors, consumers and producers, since commodities are used as primary inputs for the production of manufacturing products [4]. Thus, analyzing and accurately predicting the behavior of financial investors is of outmost importance to economic agents, policymakers and regulators and to the well-functioning of the markets.

The price bubble in crude oil and other commodities in 2007 and 2008 led some market participants to blame commodity financialization for that. Their argument being that commodity prices were exacerbated by the trading behavior of institutional investors who used centralized electronically traded futures markets resulting in increased price volatility [5]. Indeed, commodity financialization alters the mechanisms of risk-sharing, information discovery and storage through which commodity markets function [1,6,7]. However, it is not clear yet whether futures and spot price increases are attributed only to these mechanisms or to underlying economic fundamentals [8-11] or even to the influx of commodity index traders in commodity futures markets [3,12].

MWorking on this issue, Milonas & Photina [13] examined the direct effect of financialization on convenience yield (CY), an implied yield on holding inventories offering the benefit of continuing operations at times of supply scarcity. They developed a dynamic Autoregressive Distributed Lag with Exogenous Variables (ARDLX) model that captures both short and long-term impacts of the explanatory variables, produces robust empirical results and explains almost 70% of CY variability. Following Heinkel et al. [14] and Milonas & Thomadakis [15], CY is being modeled as an option to liquidate under the condition that a stock-out can occur with positive probability. As such, CY is a pivotal non-observable variable employed as a decision-making tool in commodity risk analysis. Motivated by the on-going debate on commodity financialization and the uncertainty surrounding economic phenomena, we underline the need to improve existing economic models with the sentiment and the behavioral aspects of economic agents along with the challenges that such efforts entail. The paper discusses the issue of uncertainty in financial markets and the difficulty in modeling it. Next, we turn to existing economic models that introduce behavioral aspects to model economic phenomena and the weaknesses found in addressing uncertainty. Finally, we refer to economists that introduce the sentiment of economic agents to alleviate the weaknesses in decision making models. We conclude suggesting that despite the theories presented, the mere nature of uncertainty offers limited success when analyzed with available models.

The Uncertainty in Financial Markets

Since the rise of uncertainty in 2008, it has been clear that probabilistic choice theories even when they follow a behavioral theory approach are unable to fully predict and justify the behavior of economic agents engaged in financial markets. With respect to commodity financialization, Fattouh et al. [2] argue that besides the economic reasons that justify the influx of financial investors in commodity futures markets since early 2000, this behavior could be attributed to altered investor preferences in favor of market-based instruments aiming to smooth consumption. The question that arises is about the underlying mechanism that drives financial investors to alter their preferences which results to commodity financialization.

Researchers concentrate on the end result (in this case commodity financialization) and focus on economic variables to explain the phenomenon. As a result, limited attention is given on motives that drive financial investors to change their preferences in favor of commodity futures. Furthermore, researchers rarely follow a behavioral theory approach while a neuroscience approach is totally ignored when they model choice under uncertainty. After all, investment choices are provided by centralized, electronically traded futures markets that facilitate the influx of financial investors in commodity futures markets. Available since 1992, such markets aggregate world trade information, offer limitless liquidity to participants, and assure that futures prices fully reflect all available information at any time.

Using probability, existing decision making theories can model risk but not uncertainty, since, unlike risk, uncertainty cannot be captured by a probability distribution over a set of events emerging from economic or non-economic random shocks. For most market participants economic uncertainty enters decision under risk models as a variable that measures volatility of an economic entity, such as the VIX, GDP, and exchange rates [16]. For example, the Volatility Index of the Chicago Board Options Exchange (VIX), available since 1990, is the most popular volatility measure and captures better bad events with origin in financial and stock markets [17]. VIX reflects the volatility expected by the financial market during the next 30 days and it is easily measured. Since extremely volatile data series prohibit forecasting of stock and bond prices, VIX is used as a proxy of economic uncertainty as in the paper by Milonas and Photina [13]. The paper investigates the existence of commodity financialization and VIX was found to be a significant determinant of commodity financialization. While VIX is been used extensively in pricing financial models, such models are likely to be improved if coupled with behavioral aspects of financial investors.

Modeling Choice Under Risk

Classical economists did pay attention to psychology of people. Adam Smith in “The Theory of Moral Sentiments” proposed psychological explanations for individual behavior. Furthermore, Neo-classical economists, in order to examine the decision-making process, made assumptions about the nature of economic agents that attributed to their economic behavior. Samuelson [18] followed a preference-based approach to model choice under certainty imposing rationality axioms on these preferences. However, rationality often is false for real people since many of the axioms are violated (i.e., a person cannot decide when choices have just perceivable differences, the manner of presenting alternatives matters for choice, in aggregation of preferences the order matters, and a person cannot decide when changes in taste exist). Von Neumann J and Morgenstern O [19] model choice under risk based on the expected utility theory. Besides rationality, the authors introduce four more assumptions on preferences: reduction of compound lotteries, independence of irrelevant alternatives, continuity, and monotonicity. Violations of this theory also exist including the Allais Paradox, explained either by the fact that people are not rational, or because people cannot process very small/high probabilities or because of the framing effect where equivalent descriptions of a decision problem lead to systematically different decisions. Another explanation that people often give inconsistent answers is when pressed to provide them quickly.

Kahneman and Tversky [20] address the documented violations of the expected utility theory and model decision making under uncertainty according to the prospect theory based on the following four elements: reference dependence, loss aversion (drives people to be more sensitive to losses than to gains of the same magnitude), diminishing sensitivity (makes people to become risk seeking over losses but risk averse over gains), and probability weighting (leads people to overweight unlikely extreme outcomes). The prospect theory has limitations since it predicts that people will sometimes choose dominated gambles and cannot be applied to gambles with more than two nonzero outcomes. In a new paper Tversky and Kahneman [21] presented the cumulative prospect theory to alleviate these limitations but researchers rarely implement this new theory because it is difficult to determine the reference point against which every evaluation is made in order to define gain or loss [22]. The weaknesses in modeling risk mentioned above compel us to look at other factors, such as investor sentiment, when analyzing investor choices especially and dealing with situations characterized by uncertainty. We discuss this issue in the context of commodity markets and CY.

Investor Sentiment

Financial investors’ behavior in the commodity futures market has spill-over effects on commodity physical market and consequently inventory and CY. Financial investors do not have a physical position in commodities to offset and so they do not enter the markets as hedgers. In contrast, they enter commodity futures markets as speculators and their trading contributes to increased commodity price volatility. The behavior of commodity market participants can be examined by analyzing a number of various factors. One such factor is the information on positions held by market participants available from the Commodity Futures Trading Commission, although such data might not be sufficient. Yuan et al. [23] introduced a behavioral theory approach and study the behavior of commercial and non-commercial traders in 26 commodity futures markets. In particular, they implement the probability weighting of commodity futures returns proposed by Kahneman and Tversky in the cumulative prospect theory. According to their empirical findings, commodities with a high probability weighting value significantly underperform their low value pairs by 11% per annum and also attract excess demand from non-commercial traders.

Bonato et al. [24] forecast realized price volatility of 14 important agricultural commodities using the daily agricultural commodity specific Thomson Reuters Market-Psych Indices (TRMI) to capture investor sentiment and/or realized moments like leverage, realized kurtosis, and realized jumps. They argue that investor sentiment inclusion, improves only marginally (if at all) forecasts that rely only on realized moments. This empirical finding suggests that corporate decision makers, investors, and policymakers need to concentrate only on realized moments in order to forecast future realized price volatility of agricultural commodities.

Bonato et al. [24] follow two paths of thought to justify and interpret investor sentiment insignificance. First, increased volatility in food prices is mainly attributed to rare disaster risks originated by climate change in the post Global Financial Crisis of 2007-2009. Even if investor sentiment contributes to increasing food price volatility besides biofuel production, market speculation, and increasing demand paired with decreasing food inventory, realized moments dominate since they capture better climate change risks. Second, sentiment insignificance could be attributed to linear forecasting models incapability to capture nonlinearities. Machine learning approaches able to handle “big data”, such as random forests [25], are more appropriate than linear forecasting models to uncover nonlinearities between sentiment and realized volatility, in case these nonlinearities exist. Thus, we need novel approaches to analyze commodity markets and model choice behavior.

Contemporary research on choice under risk and uncertainty, concentrates on studying and interpreting observed behavior in commodity futures markets rather than economic behavior based on assumptions set by researchers. In particular, novel indices, such as the Thomson Reuters Market-Psych Indices (TRMI), are constructed to capture investor sentiment from news articles, social media, press releases, and regulatory filings.

Concluding Remarks

When attempting To analyze uncertainty, besides including economic variables, the behavior of economic agents is important to consider in economic models. Decision making should not ignore the people’s behavior and their sentiment. Yet, the mere nature of uncertainty which cannot be described by a formal distribution with real moments, presents difficulties in utilizing methods used in risk analysis. Furthermore, the inclusion of sentiment in economic models is insufficient to enhance model prediction since real data on basic estimates are also necessary to create a prediction model. Thus, we need to uncover novel approaches to handle uncertainty in models that include the behavioral aspects of market participants.

References

1. Cheng IH, Xiong W (2014) Financialization of commodity markets. Annual Rev Financ Econom 6: 419-441.
2. Fattouh B, Kilian L, Mahadeva L (2013) The role of speculation in oil markets: What have we learned so far? The Energy J 34(3): 7-33.
3. Kupabado MM, Kaehler J (2021) Financialization, common stochastic trends, and commodity prices. J Futures Markets 41(12): 1988-2008.
4. Triantafyllou A, Bakas D, Ioakimidis M (2023) Commodity price uncertainty as a leading indicator of economic activity. International J Finance Econom 28(4): 4194-4219.
5. Masters MW (2008) Testimony Before the Committee on Homeland Security and Governmental Affairs. Retrieved from Technical Report, US Senate.
6. Tang K, Xiong W (2012) Index investment and the financialization of commodities. Financial Analysts J 68(6): 54-74.
7. Sockin M, Xiong W (2015) Informational frictions and commodity markets. J Finance 70(5): 2063-2098.
8. Yang J, Li Z, Miao H (2021) Volatility spillovers in commodity futures markets: A network approach. J Futures Markets 41(12): 1959-1987.
9. Kilian L, Murphy D (2014) The role of inventories and speculative trading in the global market for crude oil. J Appl Econometr 29(3): 454-478.
10. Juvenal L, Petrella I (2012) Speculation in the oil market. Federal Reserve Bank of St. Louis.
11. Irwin HS, Sanders RD (2012) Testing the Masters Hypothesis in commodity futures markets. Energy Econom 34(1): 256-269.
12. Basak S, Pavlova A (2016) A model of financialization of commodities. J Finance 71(4): 1511-1556.
13. Milonas NT, Photina EK (2024) The convenience yield under commodity financialization. J Futures Markets 44(4): 631-652.
14. Heinkel R, Howe ME, Hughes S (1990) Commodity convenience yields as an option profit. J Futures Markets 10(5): 519-533.
15. Milonas NT, Thomadakis SB (1997) Convenience Yield As Call Options: An Empirical Analysis. J Futures Markets 17(1): 1-15.
16. Baker S, Bloom N, Davis SJ (2016) Measuring economic policy uncertainty. Quarterly J Econom 131(4): 1593-1636.
17. Bloom N (2014) Fluctuations in uncertainty. J Economic Perspectives 28(2): 153-176.
18. Samuelson PA (1937) A note on measurement of utility. Rev Econom Studies 4(2): 155-161.
19. Von Neumann J, Morgenstern O (1947) Theory of games and economic behavior. Princeton University Press.
20. Kahneman D, Tversky A (1979) Prospect theory: An analysis of decision under risk. Econometrica 47(2): 263-292.
21. Tversky A, Kahneman D (1992) Advances in prospect theory: Cumulative representation of uncertainty. J Risk Uncertainty 5: 297-323.
22. Barberis N (2013) Thirty years of prospect theory in economics: A review and assessment. J Economic Perspectives 27(1): 173-196.
23. Yuan J, Xu Q, Wang Y (2023) Probability weighting in commodity futures markets. J Futures Markets 43(4): 516-548.
24. Bonato M, Cepni O, Gupta R, Pierdzioch C (2024) Forecasting the realized volatility of agricultural commodity prices: Does sentiment matter? J Forecasting 43(6): 2088-2125.
25. Bonato M, Çepni O, Gupta R, Pierdzioch C (2023) El Niño, La Niña, and forecastability of the realized variance of agricultural commodity prices: Evidence from a machine learning approach. J Forecasting 42(4): 785-801.