Financial Speculation Impact on Agricultural and Other Commodity Return Volatility: Implications for Sustainable Development and Food Security

Abstract

Global commodity markets, due to major health crises, political tension, sanctions, growing demand, and other global supply and demand factors, are currently particularly unstable. In addition to the macro-environmental factors that drive the prices, agricultural and other commodity markets are becoming more susceptible to the continuously-growing speculation on major commodity exchanges. Therefore, the purpose of this study is to analyze the influence of financial speculation on agricultural and other commodity prices and return volatility. In our study, we use daily returns on wheat, soybean, corn, and oats futures from the Chicago Mercantile Exchange as well as two additional commodities (crude oil and gold) to compare the extent of this effect. To measure this impact, we, besides traditional tools for time-series analysis, apply the threshold autoregressive conditional heteroskedasticity (TGARCH) technique. We also provide a model using dummy variables for the season to determine whether or not financial speculation’s impact on return volatility differs among seasons, as seasonality plays an important role in return dynamics for agriculture. Our study’s findings show that financial speculation, except for the oats market, either has no impact or makes the underlying futures returns less volatile. Therefore, we draw the conclusion that either there is no relationship between the rise in short-run speculation and the volatility of agricultural commodity prices, or the link is at best questionable. Research results provide important implications for the sustainable development of commodity markets, as passive legislation measurers can be seen as more effective ones compared to more strict active ones in order to maintain these markets liquid and capable of distributing price risks for agricultural producers and manufacturers in a challenging economic and geopolitical environment.

1. Introduction

Rapid price increases in agricultural commodities, as well as periods of increased market volatility, pose a threat to both agricultural-product consumers and producers. Since 2006, the price volatility of food grains has dramatically risen [1]. According to the Futures Industry Association, between 2007 and 2016, the volume of global commodities futures contracts surged sixfold, from one to six billion contracts [2]. This, on the one hand, provides opportunities for abnormal gains. However, the uncertainty about future prices does affect production chains, creating additional price risks. Concerns about food insecurity, which may be particularly detrimental to consumers in low- and middle-income nations, are sparked by a rise in their production costs and price volatility.

Commodity markets, due to military actions, political tension, sanctions, growing demand (due to the increase in income in individual countries), logistics problems, and other such factors, are currently particularly unstable. In addition to the macro-environmental factors that drive the prices in commodity exchanges, financial speculation may also be responsible for long-term price dynamics. Futures contracts are standardized agreements between two parties to acquire or sell a standardized asset of a certain quantity and quality at a fixed price at a future date. Commodity dealers employ them to protect themselves against price volatility. The popularity and ease of use of these financial instruments made them more appealing to speculators. Further, the fast expansion of speculation coincided with the financialization and globalization of major commodities markets [3,4,5]. The fact that the rise of financial speculation moved with the rise of commodity prices drew extra attention from researchers to whether speculation makes commodity prices more volatile.

Financial speculators can be defined as market participants who buy financial assets with the goal of profiting from changes in prices. They engage in frequent and high-risk trading activity. Speculators also participate in financial markets for purposes other than long-term investing or hedging against the price risks involved in their activities [2]. Therefore, their impact on the economy can be twofold. On the one hand, speculators provide liquidity to the markets, bring new information to the markets, and, especially in the context of commodity markets, help distribute risks [3]. On the other hand, it is not entirely clear whether financial speculators do not contribute to price bubbles in the markets, assuming that their behavior is less rational than the usual participants of these markets and thus moves asset prices away from their fundamental values explained solely through supply and demand factors, and this causes additional risks in the financial markets.

Recent empirical work on commodity futures has extensively explored futures market volatility [6,7,8]. Besides such factors as GDP growth, industry production, inflation, money supply [9], energy prices [10], supply and demand shocks, food stocks and trade restrictions [11], global supply and consumption [12], the cost of fertilizers, and major political, financial, and natural events [13], the authors also argue what additional factors may have an impact on agricultural commodity prices and volatility. According to various analysts, agricultural commodity pricing is subject to price spikes and volatility due to speculative volumes and other factors unrelated to agricultural risk hedging [14,15,16]. The over-speculation of agricultural commodities outperforms typical commercial investments in terms of hedging against underlying price risk [17]. As a result, many argue that speculation is to blame and advocate for stricter regulation, such as transaction fees, marginal account requirements, and position restrictions [18,19,20].

The aim of the research is to evaluate the relationships between financial speculation and returns from commodity futures in agriculture and other industries and to provide implications for the sustainable development of agricultural markets.

There are five primary sections to this research. The authors have exposed the relationships between financial instruments and changes in commodity prices in this section, along with the tools used to assess these relationships and a brief explanation of the theoretical basis. The authors outline their research strategies and the data they utilized in Section 2. The findings of the empirical investigation are presented in Section 3. The last parts of this paper are the discussion and broad conclusions that come from evaluating empirical research and the scientific literature.

Many factors contribute to commodity price volatility from both demand and supply; the linkages between these and speculation are not clear. In addition, new derivatives are being developed, futures trading is becoming more accessible, and index funds help investors diversify their holdings. According to the traditional theory of economic stability and speculation, the main conditions that make a market attractive to speculation are low carrying costs, almost perfect or semi-perfect competition, products traded being standardized, durable and long-lasting, and cheap enough to transport and have continuous demand [21]. In the case of futures contracts, all these conditions are valid, and therefore, speculation is frequent and its role in making these markets more efficient becomes more important. Therefore, many propositions on how to regulate speculators have arisen. Among the measures advocated are the restriction of investment positions, the increase in transaction charges, and the enhancement of market transparency [19,20].

The link between financial speculation and commodity prices has long been debated [22,23]. The first researchers who analyzed these linkages emphasized their role in commodity markets [21,24]. The role of speculators in these markets is to provide the necessary liquidity. However, an amount of speculation larger than needed to offset the long-short commercial positions in the commodity futures market is described as excessive speculation. Working’s T index is traditionally used to estimate the amount of excessive speculation [25]. Later authors focused on commodity exchanges and how trading activity affects price volatility but found mixed effects or feedback relationships. For example, in grain futures markets, returns result in net position adjustments for both commercial traders and investors [26]. Therefore, speculators are important for risk sharing and bringing new information [27,28]. For example, Borgards and Czudaj [29] draw the conclusion that long–short speculators possess significant, unique knowledge that cannot be duplicated by looking at their trading activity with an eight-trading-day time lag. However, the outlook on speculation has changed since the 2007 crisis. This gave rise to the Master’s (2008) [30] hypothesis, which asserts that the destabilization of major commodity markets was caused by the excessive growth of index investors since the early 2000s. Additionally, these negative trends moved from the futures market to the spot market, driving up the price of food around the world [20]. This becomes especially important when using cash settled futures contracts. For example, because financial investors do not physically purchase crude oil, the public began to think that their activities, which were quickly expanding, were driving up oil price volatility and producing the perplexing dynamic co-movements [31]. In the years that followed, various researchers thoroughly examined this hypothesis, but came up with mixed results, as speculation is found to be more often a result of increased market volatility [32,33]. Later on, authors emphasized crisis periods by comparing them to regular periods of market activity [34,35,36]. Some research on the non-US markets found positive (destabilizing) effects [14,37]. Some authors even observe stabilizing influences or reverse effects [6,19]. However, some found destabilizing influences, most notably when examining less liquid or smaller commodity markets [22,37]. Agricultural commodity price speculation, according to these studies, jeopardizes food security and market stability. Other issues to be concerned about include agricultural and agribusiness hazards, income stability, and planning choices. This is especially significant given the current pandemic situation, which has influenced commodity price dynamics and is the next large crisis after the 2007/2008 economic turmoil [38,39,40]. However, little study has been conducted on this time period and how speculation interacts with commodity prices [41]. In contrast, recent COVID-19 period research has mostly focused on market co-movements [42,43].

Long-term and short-term speculation are the two main kinds of indicators used in research on speculation [22]. Long-run speculation is described by the number of non-commercial positions held, whereas short-run speculation is described by trading activity. Two economic theories that investigate the impact of speculation on financial market prices are effective market hypotheses [44,45,46,47] and behavioral theories [48,49,50]. On the one hand, speculators provide new information to commodity markets; on the other hand, they have different motivations than conventional corporate users who hedge against price risk, and their behavior patterns may lead to commodity prices varying from their underlying fundamental value. Herd behavior is emphasized in some theoretical research on commodity markets [51,52,53]. When investigating the involvement of speculators in futures markets, the storage cost model [54,55] is often used to explain price volatility and the discrepancies between spot and futures prices using inventory numbers, interest rates, and desired profitability. According to the hedging pressure theory [3,56], which is an extension of the normal backwardation theory [24], futures market speculators obtain risk premiums by taking on risk, modeling prices on risk premiums, and employing parity models. As a consequence, non-commercial product market positions are required for market liquidity.

Uncertainty about the possibility and scale of pandemics is also fueling commodity price instability. In recent research, much attention has been paid to fundamental factors or macro-environmental factors that have an impact on commodity prices during the pandemic period. The COVID-19 pandemic has had a significant impact on the world’s quality of life, political stability, environmental sustainability, and global economy [57]. As for product markets, researchers identify such threats as interference in supply chains, food security, food fraud, and globalization [58], in addition to food security anxiety and policy implications [59]. In addition, restrictions on import and export played an important role in agricultural product formation during this time period [60]. Financial sectors and financial markets are highlighted in recent work on the effects of pandemics, in addition to other effects on the economy [61]. The study’s empirical findings showed that although oil prices rose in tandem with financial success, a growth in the number of COVID-19 registered patients had an unprecedentedly negative impact on financial development [62]. In addition, the agricultural product markets have been strongly affected by the increase in the prices of energy and other resources.

Various methodological frameworks have been used to investigate the interplay between the financial markets, speculation in derivatives, and the agricultural commodities markets. This includes methods used for time-series analysis: the Johansen cointegration test [63], the Granger non-causality test [63,64], vector autoregression (VAR) models [65], nonparametric regression tests [65], the autoregressive distributed lag (ARDL) model [66], the generalized autoregressive conditional heteroskedasticity (GARCH) model [19,50], panel Granger non-causality analysis [22], the vector error correction model (VECM) [63,67], the dynamic conditional correlation (DCC) GARCH model [68,69], the stochastic volatility (SV) model [70,71], the standard heterogeneous autoregressive (HAR) model [72,73], the structural vector autoregressive (SVAR) model [1,34], the continuous Granger non-causality test [74], and quantile regression models [75,76].

Previous studies used a more traditional approach to modeling speculation in commodity markets. However, the models offered do not account for seasonality, which is common in agricultural markets [77,78]. The primary goal of this study is to strengthen the impact of speculation on return volatility by expanding other authors’ agricultural product price volatility models. Using theoretical and empirical derivatives speculation theories, we study the influence of derivatives speculation on global commodity prices. Many studies [16,37] only look at agricultural products and not at others. For example, metals, especially gold, are considered a “safe haven” for investors [79]. Therefore, this paper focuses on all three commodity groups—agriculture, metals, and energy—to see how financial speculation affects different product groups. As other authors concluded, oil changed its role from serving as a safe haven for agricultural and metal commodities, but lost that function after the global financial turmoil [80]. Therefore, this paper also makes a distinction between the time before and after a crisis, focusing on the 2020 COVID-19 pandemic in particular. This time period has yet to be sufficiently analyzed in terms of financial speculation’s impact on commodity returns, but other authors emphasize that, in general, pandemic turmoil has a great impact on the cross-correlation of commodity future markets and leads to a cross-market spillover of financial risks while altering the hedging abilities of these commodities [81].

Given the recent challenges related to financial speculation and rising food prices, this paper looks into the relationship between financial speculation and returns from commodity futures to see if financial speculation makes commodity prices more volatile. This paper contributes to the work of other authors and proposes a new methodology that analyzes the impact of financial speculation on return variability in the context of seasonality. Seasonality is an important characteristic of agricultural markets, which are the main focus of this work. The data analyzed in the study also include the most recent periods; the study separately analyzed the pandemic period, which has not been analyzed yet in the works of other authors.

2. Materials and Methods

2.1. Variables Used in the Study

To investigate links between price, returns, and speculation and to test whether short-run speculation drives the return volatility of agricultural and other commodity futures contracts, we use the generalized autoregressive conditional heteroskedasticity models (GARCH) along with typical instruments for time-series analysis such as the unit root-test. The dependent variables are the continuous futures price or returns from futures contracts. The short-term speculation index is the independent variable. To calculate it, we divide the total trade volume (TV) by the open interest (OI).

First, we define the variables for each of the models. To begin, we generate a return from the commodity futures series by using the natural log of future contract prices. The period-to-period logarithmic difference in futures contract pricing t and t-1 is used to calculate returns on commodity futures and can be seen in Formula (1). Price indices such as the opening, closing, lowest, and highest prices may be used in analysis of futures markets to determine the underlying relationships. We have decided to go with the continuous futures contract day’s closing price. To display it as a percentage, as in previous research [78], we multiply this by 100. The futures contract’s return is our research dependent variable, represented by $R_r$, and will be used for all commodities in the study.

$$R_r=\ln \left(\frac{P_t}{P_{t-1}}\right)\times 100$$

(1)

where: $R_t$ is the return on a futures contract; $P_t$ is the futures contract price; $t$ is the time period; and $\ln$ is the natural log.

Next, the TV/OI is a short-term speculation index that measures speculation in the commodity markets and is described in Formula (2). This speculation index is based on daily data and is applied to futures exchanges that do not provide detailed information on positions, such as commercial and non-commercial long positions, or when high frequency data are needed and are used by other authors [6]. Trade volume reflects the level of speculative activity, whereas open positions represent the level of hedging activity [37]. Another term for this type of speculation is the ratio of trade volume to the total open interest (TV/OI). According to Shear [14], speculative activity has an impact on the volume of daily trade because traders who engage in speculation have a limited time horizon on a daily basis. This speculation indicator allows us to gather more data as we use it daily instead of on a weekly basis, as is the case with long-run speculation indicators, and this is especially useful when analyzing the post-2020 time period to increase the number of observations.

$${\mathrm{S}}_t=\frac{T{V}_t}{O{I}_t}$$

(2)

where: ${\mathrm{S}}_t$ is the short-term speculation index; $T{V}_t$ is futures contract trade volume; $O{I}_t$ is futures contract open interest; and $t$ is the time period.

2.2. Test for Unit-Root

We then go over the specifics of our econometric model. First, time series are tested to see if they are stationary. When assessing time series, it is essential that their statistical characteristics, namely mean, autocorrelation, and variance, stay consistent. The Augmented Dickey–Fuller (ADF) test is used to determine whether the time series used in this study are stationary. To identify whether time series have a unit root that characterizes non-stationary processes and whether they have a stochastic trend, this technique incorporates a single root test [63]. The ADF test is built on a model described in Formula (3) [82]. We use this test to validate hypotheses—H0: unit root is present; H1: alternative hypothesis. The statistical hypothesis in this case is H0: the time series has a single root $\phi =0$. A random walk is described by a time series with a parameter $\phi$ value of 0. If this parameter of the return time series does not change over time and stays statistically insignificant, the investigated cause-and-effect relationship should stay the same for the whole sample period.

$$\Delta Y_t=\phi Y_{t-1}+u_t$$

(3)

where: $Y_t$ is the dependent variable; $\phi$ is a model parameter; $u_t$ is the residual error; $\Delta$ is the first order change; and $t$ is the time period.

The extended ADF test described in Formula (4) uses a constant, a trend, and a larger number of time lags [83]. This allows us to determine whether the time series is stationary by taking into account the long-term trend. Indicators such as prices are anticipated to increase over time in this respect. As a consequence, long-term economic expansion or the rise of agricultural markets that result in long-term changes in pricing are corrected.

$$\Delta Y_t=\alpha +\beta t+\phi Y_{t-1}+{{\displaystyle \sum }}_{i=1}^j{\theta }_i\Delta Y_{t-i}+u_t$$

(4)

where: $Y_t$ is the dependent variable; $\alpha$, $\beta$, $\phi$, ${\theta }_i$ are the model parameters; $u_t$ is the residual error; $\Delta$ is the first order change; $i$ are time lags; $j$ is the number of time lags; and $t$ is the time period.

The number of time lags is determined by taking the third-degree root of the sample size, thus $j=\sqrt[3]{n}$. Next, the hypotheses for the ADF test can be described as:

H0: $\phi =0$. Time series that have a unit root.

H1: $\phi <0$. Time series that do not have a unit root.

Next, after providing descriptive statistics and results of the unit root test, we use the ARCH models to model returns from the future contracts and to assess the impact of speculation on the conditional return volatility. These models have been widely used to analyze price and return volatility in financial markets. In ARCH modeling, the variance of the current error term is given as a function of the squared past error terms [8].

2.3. Month Selection Model

We assume that returns from futures contracts are stationary processes but have non-constant conditional variance. Therefore, we employ ARCH modeling to examine how speculative variables influence price or return conditional volatility. In ARCH modeling [66], stationary time series are required for reliable parameter estimates. Consequently, if the price has a unit-root, we use returns, which, as mentioned before, are first-order logarithmic price differences. Engle (1982) [84] proposed initial ARCH models, and Bollerslev (1986) [85] improved the methodology and created the generalized GARCH approach. GARCH models are widely used in the study of financial markets because the data used in time series models are usually characterized by a non-uniform distribution of errors. In other words, returns from financial assets are characterized by clustering, i.e., prior variability explains subsequent variability. Therefore, these methods allow for efficient analysis, evaluation, and forecasting of returns from financial assets and are used in the works of other researchers. The model enables the estimation of the effect of speculation on the variability of returns. Because of this, the study’s results can be compared to what other researchers have found. Most other studies employ the generalized GARCH approach and use one residual error and one generalized variability time offset, GARCH (1,1) [16,76,78]. This reduces the number of parameters, makes them easier to read, and improves the economic interpretation of the results. Similarly to other authors, our model employs a one-day offset and uses the GARCH (1,1).

GARCH models are made up of two parts: mean and variance equations. The mean equation includes an autoregressive equation of future contract returns. The second equation of the model, known as the equation of variation, allows for the evaluation of the time-lagged links between return volatility and the previous period residual from the mean equation (ARCH effect), and the generalized autoregressive volatility (GARCH effect). The residual error depicts the influence of innovation on return volatility, whereas the preceding generalized volatility parameter indicates whether or not the volatility is clustered. External (exogenous) variables can be put into both of these equations. As we primarily focus on whether speculation has an impact on return volatility, we include the speculation index in our variance equation.

We also add seasonality to our model. To select the month with the highest return volatility for each commodity, we provide a preliminary GARCH (1,1) in which months are included in the variance computation to examine seasonal volatility. To avoid multicollinearity, we only used 11 months from January to November because agricultural prices are less volatile during the winter months and other authors also excluded these months [8,77,78,86,87]. Spring can be assumed to be the most volatile season for agricultural futures markets as most new information on crops and harvests reaches the market. The following Formulas (5) and (6) describe our suggested month-selection model:

Mean equation:

$$R_t={\alpha }_0+{\alpha }_1{R}_{t-1}+u_t$$

(5)

Variance equation:

$$h_t^2={\beta }_0+{\beta }_1{u}_{t-1}^2+{\beta }_2{h}_{t-1}^2+{\displaystyle \sum }_{i=1}^n{\gamma }_i{D}_i$$

(6)

where: returns $R_t$ in the mean equation with parameters ${\alpha }_0$, ${\alpha }_1$, and a residual error $u_t$ with a variance of $h_t^2$. In the variance equation ${\beta }_0$ is the constant; ${\beta }_1{u}_{t-1}^2$ is the innovation; ${\beta }_2{h}_{t-1}^2$ is the previous volatility;$t$ is the time period; ${\gamma }_i{D}_i$ is the month’s effect on return volatility; $i$ is the index for the month; and $n$ is the number of months excluding December.

2.4. Main TGARCH Model

In addition to the GARCH model, researchers also apply many extensions to this model, such as EGARCH, APGARCH, etc. This is where TGARCH stands out. Its parameters are easier to give economic interpretations and are relevant in the context of the impact of speculation on returns, as it analyzes whether positive or negative information has a symmetric effect on return variability by explaining behavioral patterns and whether market participants are less or more rational. Compared to classical GARCH, this model is supplemented with an asymmetry coefficient in the equation of variation, which allows for a better estimate of the effect of new information entering the market on the variability of prices or returns and estimates asymmetric relationships that are characteristic of market participants, assuming that market participants do not necessarily behave rationally and tend to overestimate negative information.

Therefore, besides the classical GARCH model, we also use an extended version, the Zakoian’s (1994) [88] threshold autoregressive conditional heteroskedasticity (TGARCH) model. In contrast to the traditional GARCH model, a binary factor is added to the variance equation to calculate the influence of negative returns on volatility and the asymmetry of the connection [89]. This is especially useful in the analysis of financial data as a positive shock is implied by good news and a negative shock is implied by bad news. In our results section, we provide estimations for both the GARCH and TGARCH models. The TGARCH model with the speculation index in the variance equation is identical to the GARCH model and is described in Formulas (7) and (8), but unlike the GARCH model, it has the component ${\beta }_3{u}_{t-1}^2{d}_{t-1}$ in its variance equation that measures if negative returns have an asymmetric effect on further return volatility.

Mean equation:

$$R_t={\alpha }_0+{\alpha }_1{R}_{t-1}+u_t$$

(7)

Variance equation:

$$h_t^2={\beta }_0+{\beta }_1{u}_{t-1}^2+{\beta }_2{h}_{t-1}^2+{\beta }_3{u}_{t-1}^2{d}_{t-1}+{\beta }_4{S}_{t-1}$$

(8)

where: returns $R_t$ are in the mean equation with parameters ${\alpha }_0$, ${\alpha }_1$ and a residual error $u_t$ with a variance of $h^2$. In the variance equation ${\beta }_0$ is the constant; ${\beta }_1{u}_{t-1}^2$ is the innovation; ${\beta }_2{h}_{t-1}^2$ is the previous volatility; ${\beta }_3{u}_{t-1}^2{d}_{t-1}$ is the asymmetrical element; and the dummy variable $d_t=1$ if $u_{t-1}<0$ and $d_t=0$ otherwise. If ${\beta }_3\ne 0$, then an asymmetric relationship exists. Next, ${\beta }_4{S}_{t-1}$ is used to assess the impact of short-run speculation on return volatility; $t$ is the time period.

2.5. TGARCH Model with Seasonality

Then, using a one-period autoregressive offset TGARCH (1,1), we define our proposed model with seasonality. Here, an extra variable is used to model financial speculation as a season-weighted variable consisting of the speculation index multiplied with a seasonal dummy variable. This model can be used to calculate the impact of seasonally weighted financial speculation on return volatility and its direction. The mean equation given in Formula (9) includes an autoregressive equation of future contract return. The variance equation given in Formula (10) allows for the evaluation of the autoregressive link between return variability and external (exogenous) variables, in this case, three exogenous variables. In the variance equation, we use the month that, after a preliminary test for seasonality, is found to have the most return volatility. We name the GARCH and TGARCH models without seasonality “Approach I” and the GARCH and TGARCH models with seasonality “Approach II.” Approach I only employs the speculation index, whereas Approach II employs the speculation index as well as the month effect and the multiple effect of speculation and month.

Mean equation:

$$R_t={\alpha }_0+{\alpha }_1{R}_{t-1}+u_t$$

(9)

Variance equation:

$$h_t^2={\beta }_0+{\beta }_1{u}_{t-1}^2+{\beta }_2{h}_{t-1}^2+{\beta }_3{u}_{t-1}^2{d}_{t-1}+{\beta }_4{S}_{t-1}+{\beta }_5{D}_{t-1}+{\beta }_6{D}_{t-1}{S}_{t-1}$$

(10)

where: ${\beta }_6{D}_{t-1}{S}_{t-1}$ is the multiple effect of the speculation index $S_{t-1}$ and the month’s effect $D_{t-1}$. For other parameters’ descriptions, see Formulas (7) and (8).

Then we look to see if the estimated p-values of the GARCH/TGARCH model parameters are less than 0.05, indicating strong clustering of volatility, exogenous variables, and so on. To compare models, information criteria are applied: the Hannan–Quinn information criterion (HQC), the Akaike information criterion (AIC), and the Schwartz information criterion (BIC). A comparison of different models allows one to determine which one of the models is more effective and gives better results on speculation impact on return volatility. Therefore, in the results section, models are presented and compared with each other to evaluate which model has lower information criteria values, taking into account whether the results of the models are similar or whether the results of the newly constructed model do not deviate much from the classical model. The newly proposed model, which includes months and seasonality, extends the research carried out by other authors by focusing more on the characteristics of seasonality in the agricultural sector and determining whether speculation amplifies return volatility during more volatile months.

2.6. Data

We investigate the association between price movements and speculative trading behavior. More specifically, we explore how futures speculation affects futures contract return volatility by using four regularly traded agricultural and two metal/energy commodities. Wheat, soybeans, corn, and oats futures are chosen as agricultural commodities; crude oil futures are chosen as energy commodities; and gold futures are chosen as metal commodities. The Chicago Board of Trade deals with these agricultural futures, whereas the New York Mercantile Exchange deals in gold and crude oil futures. The data platforms Bloomberg and Barchart give us daily continuous futures prices, total open interest, and trade volume. The continuous future contract is a dynamic contract that changes at the beginning of each contract month’s trading day. Using this information, we determine the ratio of each commodity market’s short-term speculation. From 15 January 1986, until 21 September 2021, we gathered data on a daily basis. Throughout our investigation, we saw significant price changes and, especially in crude oil and gold markets, large increases in speculative activity (see Appendix A and Appendix B). The sample is then separated into two subsamples: the full sample and after 2020. Price trends and shocks induced by the epidemic are shown in the post-2020 era.

3. Results

We begin with descriptive statistics of all six commodities from the US markets: wheat, soybeans, corn, oats, gold, and oil futures (see

Table 1. Descriptive statistics of agricultural and other commodity futures.

Indicator	Wheat		Soybean		Corn		Oats		Gold		Oil
	Full	2020+	Full	2020+	Full	2020+	Full	2020+	Full	2020+	Full	2020+
Price
Average	445.0	603.2	809.0	1134	331.1	451.2	212.8	338.5	756.4	1789	44.59	50.12
Minimum	224.0	475.8	410.0	821.8	142.8	312.0	93.75	253.5	253.7	1478	10.42	11.57
Maximum	1283	775.0	1768	1643	838.8	732.3	551.5	551.5	2065	2069	145.3	75.25
STDEV	163.5	73.17	294.8	243.5	138.6	110.2	87.82	67.41	501.3	113.9	28.79	15.14
STDEV, %	36.73	12.13	36.44	21.47	41.87	24.42	41.28	19.91	66.28	6.37	64.56	30.21
Skewness	1.121	0.263	0.870	0.178	1.360	0.495	0.775	1.064	0.756	−0.430	0.839	−0.233
Kurtosis	1.055	−1.016	−0.164	−1.499	1.412	−0.932	−0.252	0.541	−0.907	−0.079	−0.360	−0.781
Return
Average	0.008	0.049	0.010	0.066	0.008	0.066	0.015	0.139	0.018	0.036	0.012	0.033
Minimum	−21.34	−4.19	−25.02	−6.98	−27.40	−11.37	−26.84	−8.80	−9.81	−5.114	−56.86	−56.86
Maximum	9.054	5.129	7.806	6.382	9.601	7.039	12.11	5.500	8.887	5.775	25.06	25.06
STDEV	1.796	1.685	1.489	1.352	1.666	1.818	2.056	1.716	1.019	1.220	2.580	5.495
Skewness	−0.361	0.464	−1.421	−0.632	−0.822	−0.545	−0.668	−0.507	−0.255	−0.374	−1.620	−2.671
Kurtosis	6.939	0.380	18.76	5.807	14.52	5.499	8.201	1.992	7.123	4.281	43.20	33.05
Speculation index
Average	0.372	0.321	0.507	0.271	0.301	0.227	0.221	0.117	0.316	0.511	0.366	0.468
Minimum	0.001	0.135	0.003	0.102	0.003	0.074	0.001	0.020	0.001	0.166	0.001	0.123
Maximum	2.000	0.682	2.229	0.517	1.452	0.584	2.000	0.402	2.677	1.214	1.820	1.820
STDEV	0.184	0.102	0.242	0.077	0.143	0.079	0.183	0.063	0.188	0.187	0.147	0.194
Skewness	1.568	0.763	1.087	0.846	1.378	1.089	2.323	1.117	1.490	1.204	1.248	2.603
Kurtosis	5.926	0.277	2.190	0.324	3.203	2.182	10.34	1.779	5.172	1.601	4.567	11.96

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021.

). We first look at descriptive statistics from the full sample. Note that prices for wheat, soybeans, corn, and oats are given in US dollars per 100 bushels, whereas the price for gold is given in US dollars per ounce, and the price for oil is given in US dollars per barrel. The prices are more volatile for gold and oil, whereas other commodities, especially wheat and corn, have less price volatility. Crude oil futures contracts have the most volatile returns because their standard deviation is the greatest (estimated value is 2.580). Gold futures have the lowest standard deviation (1.019), whereas agricultural commodities are in between oil and gold. When only agricultural goods are taken into account, the standard deviation of returns is lowest (1.489) for soybeans and highest (2.056) for oats. This demonstrates that agricultural commodities have lower price risks than oil. Another important observation is that returns from all commodities have negative skewness, and it is close to zero, indicating that there are more small positive returns and fewer but larger negative returns. Another important indicator used in descriptive statistics for time series is kurtosis, and it is highest for oil, indicating that most returns are close to zero. Returns from wheat futures contracts, for example, have the lowest kurtosis value. The average short-term speculation index value for oats is the lowest and the highest for soybean futures. This shows that even though there is more speculation in some agricultural markets, their prices and returns are less volatile if compared to the crude oil market. When analyzing the post-2020 time period, oil has the most volatile prices and returns. Speculation is more intense in the oil and gold markets. In agricultural markets, corn has the highest volatility in prices and returns. However, speculation is greatest in the wheat market. Speculation is the least active in the oats market. Returns are the least volatile in soybean markets.

The results of the Augmented Dickey–Fuller (ADF) test are then shown using two models: one that only uses the constant and the other that uses both the constant and the trend (see

Table 2. Augmented Dickey–Fuller test results, p-values.

Indicator	Wheat		Soybean		Corn
	Full	2020+	Full	2020+	Full	2020+
Price of commodity
test with a constant	0.0706	0.6244	0.2777	0.8108	0.1615	0.8112
with a constant and a trend	0.0325	0.0707	0.1352	0.8852	0.0932	0.7687
Return of commodity
test with a constant	<0.0001	<0.0001	<0.0001	<0.0001	0.0001	<0.0001
with a constant and a trend	<0.0001	<0.0001	<0.0001	<0.0001	0.0001	0.0003
Speculation Index of commodity
test with a constant	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
with a constant and a trend	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
Indicator	Oats		Gold		Oil
	Full	2020+	Full	2020+	Full	2020+
Price of commodity
test with a constant	0.5477	0.9919	0.9693	0.0828	0.2910	0.8248
with a constant and a trend	0.1414	0.8592	0.7609	0.5930	0.2214	0.0977
Return of commodity
test with a constant	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
with a constant and a trend	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
Speculation Index of commodity
test with a constant	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	0.0057
with a constant and a trend	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	0.0015

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021.

). Estimated p-values for price time series are greater than 0.05 for all six commodities and both ADF models, indicating that these time series have a unit root and are non-stationary, with the exception of wheat, using the test with constant and trend (p-value is estimated to be 0.0325). The prices of all commodities increase over time due to economic factors, and this ends up as a non-stationary process. All other time series variables are stationary, which means that their mean, variance, and auto-correlation are stationary. Therefore, further GARCH modeling can be done without convergence errors caused by using time series that do have a unit root. Similar findings are shown by the ADF test analysis for the post-2020 time frame.

The GARCH modeling results are then examined using a model with the speculation index as an exogenous variable in the variance equation to see if speculation has an effect on return conditional volatility. To begin, we look at the results of the classic GARCH model to see if there are any differences (see

Table 3. Parameter estimates of the generalized autoregressive conditional heteroskedasticity (GARCH) model using Approach I.

Indicator	Wheat	Soybean	Corn	Oats	Gold	Oil
Mean equation:
Constant	−0.0049	0.0059	0.0245	−0.0023	−0.0003	0.0408 **
Return	0.0056	−0.0076	0.0246 *	0.0781 **	−0.0155	−0.0097
Variance equation:
Constant	0.0265	0.0308 **	0.0306 *	0.1576 **	0.0008	0.0275
Residual	0.0373 **	0.0690 **	0.0669 **	0.1001 **	0.0402 **	0.0962 **
Volatility	0.9508 **	0.9254 **	0.9224 **	0.8467 **	0.9584 **	0.8948 **
Speculation index	0.0343	−0.0151	0.0232	0.3797 **	0.0086	0.1235 *
Llik:	−17,565	−15,458	−16,487	−18,702	−11,940	−19,215
BIC:	35,184	30,970	33,028	37,459	23,935	38,485
AIC:	35,142	30,927	32,985	37,416	23,893	38,442
HQC:	35,156	30,942	33,000	37,431	23,907	38,457

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.1 are flagged with one asterisk (*), and those with a p-value less than 0.05 are flagged with two asterisks (**).

). Here, we investigate the entire time period. First, we look at the mean equation, which only has a statistically significant constant for crude oil futures contracts (estimated to be 0.0408). The parameters for return autoregression in corn and oats are statistically significant and positive. Next, we look at the variance equation. Under a p-value of 0.05, residual and volatility effects are statistically significant for all six commodities. The presence of significant volatility indicates that the return volatility of all commodities is clustered. Volatility closely resembles its lagged values due to statistically significant residual parameters. Only in the case of oats and oil futures does short-term speculation have a statistically significant and positive effect (parameters are estimated to be 0.3797 for corn and 0.1235 for oil). Therefore, this effect is stronger in the oats futures market than in oil. The smallest information criteria values are for gold, whereas the largest are for oil.

The TGARCH model is then used to look at the same time series with an extra dummy variable that shows if negative past returns affect conditional volatility in an asymmetric way (see

Table 4. Parameter estimates of the threshold autoregressive conditional heteroskedasticity (TGARCH) model using Approach I.

Indicator	Wheat	Soybean	Corn	Oats	Gold	Oil
Mean equation:
Constant	0.0080	0.0325 **	0.0193 **	0.0030	0.0185 **	0.0058
Return	0.0070	−0.0095	0.0339 **	0.0739 **	−0.0154	−0.0050 **
Variance equation:
Constant	0.0577 **	0.0206 **	0.0401 **	0.1670 **	0.0057 **	0.0533 **
Residual	0.0519 **	0.0745 **	0.0785 **	0.1055 **	0.0533 **	0.0912 **
Volatility	0.9420 **	0.9383 **	0.9265 **	0.8715 **	0.9544 **	0.9185 **
Asymmetry factor	−0.4197 **	−0.2807 **	−0.0483	−0.1679 **	−0.3349 **	0.2952 **
Speculation index	0.0013	−0.0090	0.0056	0.1826 **	0.0055	0.0565
Llik:	−17,513	−15,348	−16,426	−18628	−11922	−19,190
BIC:	35,090	30,760	32,916	37,320	23,907	38,444
AIC:	35,040	30,710	32,866	37,270	23,858	38,394
HQC:	35,057	30,727	32,883	37,287	23,875	38,411

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.05 are flagged with two asterisks (**).

). More parameters in the mean equation are statistically significant than in the previous GARCH model. Most notably, crude oil has a negative autoregression parameter (it is estimated to be −0.0050) but no statistically significant constant. However, there are statistically significant constants in soybean, corn, and gold. If we look at the variance equation, both the residual and volatility parameters are statistically significant, as they were in the previous GARCH model. The constant is statistically significant and positive for all six commodities. Except for corn, the asymmetry factor is statistically significant in all cases. Except for oil, all parameter values are negative, indicating that negative changes in returns are followed by less volatility than vice versa. Oats futures are the only ones where short-term speculation has a statistically significant and positive effect (calculated at 0.1826). The smallest information criterion values are for gold, whereas the largest are for oil. If we compare this with the previous GARCH model, information criteria are smaller in this model for all six commodities.

Next, we provide the estimation results for the season selection model (see

Table 5. Parameter estimates of the GARCH model for seasonal volatility patterns.

Month	Wheat	Soybean	Corn	Oats	Gold	Oil
January	0.0038	−0.0093	−0.0415	−0.1570 *	0.0050	0.0029
February	0.0759 **	−0.0132	−0.0165	−0.0854	0.0015	0.0219
March	0.0807 **	0.0199	0.0409 *	0.0297	−0.0035	0.0316
April	0.0050	−0.0097	−0.0219	0.0260	−0.0024	−0.0196
May	0.1129 *	0.0616 **	0.0500	−0.0088	−0.0004	0.0122
June	0.0275	0.1862 **	0.1893 **	0.3730 **	0.0020	0.0171
July	−0.0276	−0.0525 **	−0.0863 **	−0.0931	−0.0017	0.0042
August	0.0065	0.0012	−0.0098	−0.0688	−0.0042	0.0254
September	0.0384 *	0.0100	0.0016	−0.0647	0.0030	0.0066
October	−0.0149	−0.0058	−0.0336	−0.1302	−0.0017	−0.0021
November	0.0306	−0.0104	0.0173	−0.0571	−0.0010	0.0709

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.1 are flagged with one asterisk (*), and those with a p-value less than 0.05 are flagged with two asterisks (**).

). Here, only parameters from the variance equation from the classic GARCH model are displayed. Note that December is removed to prevent multicollinearity. We look at these six commodity markets when the returns from their futures contracts are the most volatile. Returns on soybeans, corn, and oats are most volatile in June, and the effect is statistically significant. Wheat returns are most volatile in May and statistically significant at a p-value above 0.05. Neither gold nor oil have a statistically significant effect, indicating that these commodities are hardly affected by seasonality, in contrast to agricultural futures. If we look at their parameter values, the gold futures market is the most volatile during November, whereas the oil market is the most volatile during January. As a result, we chose June for soybeans, corn, and oats; May for wheat; January for gold; and November for oil for further analysis.

Following the selection of months for each commodity, the GARCH modeling results are investigated using two extra exogenous variables in the variance equation: the month dummy variable and its multiplication with the speculation index (see

Table 6. Parameter estimates of the GARCH model using Approach II.

Indicator	Wheat	Soybean	Corn	Oats	Gold	Oil
Mean equation:
Constant	−0.0034	0.0144	0.0294 *	0.0032	−0.0005	0.0395 **
Return	0.0049	−0.0189 *	0.0225 *	0.0777 **	−0.0148	−0.0098
Variance equation:
Constant	0.0456 *	0.0281 **	0.0452 **	0.1911 **	−0.0001	0.0332
Residual	0.0401 **	0.0650 **	0.0530 **	0.0954 **	0.0407 **	0.0961 **
Volatility	0.9420 **	0.9161 **	0.9289 **	0.8346 **	0.9572 **	0.8948 **
Speculation index (1)	0.0062	0.0009	−0.0292	0.3596 **	0.0119	0.0962
Month (2)	0.1279	0.5053	0.1622	0.2299	0.0207 *	−0.0726
1 × 2	0.0101	−0.4322	0.1112	0.9024	−0.0400	0.3633
Llik:	−17,537	−15,325	−16,421	−18,669	−11,935	−19,212
BIC:	35,146	30,723	32,915	37,411	23,943	38,496
AIC:	35,089	30,666	32,858	37,354	23,886	38,440
HQC:	35,109	30,686	32,877	37,373	23,906	38,459

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.1 are flagged with one asterisk (*), and those with a p-value less than 0.05 are flagged with two asterisks (**).

). The parameter estimations in the mean equation are similar to the previous model. However, in this case, the constant for corn is also statistically significant (it is estimated to be 0.0294), and the autoregression parameter for soybean is statistically significant but negative (it is estimated to be −0.0189). Under a p-value of 0.05, residual and volatility are statistically significant for all six commodities. The constant is statistically significant in all cases but gold and oil. Only in the case of oats does short-term speculation have a statistically significant and positive effect (estimated to be 0.3596). The month has a small but statistically significant effect on gold (estimated to be 0.0207). The month’s effect multiplied by speculation is not statistically significant in any case. The smallest information criterion is for gold, whereas the largest is for oil. If compared to the previous GARCH model, this model has the smallest information criteria for all commodities except for oil, showing that it better describes the returns from futures contracts.

The TGARCH model is then used to investigate as previously, but using three exogenous variables instead of one (see

Table 7. Parameter estimates of the TGARCH model using Approach II.

Indicator	Wheat	Soybean	Corn	Oats	Gold	Oil
Mean equation:
Constant	0.0100	0.0366 **	0.0266 **	0.0056	0.0192 **	0.0056
Return	0.0059	−0.0186 *	0.0312 **	0.0730 **	−0.0144 **	−0.0049
Variance equation:
Constant	0.0672 **	0.0247 **	0.0542 **	0.2011 **	0.0047 **	0.0573 **
Residual	0.0515 **	0.0680 **	0.0639 **	0.1040 **	0.0543 **	0.0918 **
Volatility	0.9396 **	0.9364 **	0.9335 **	0.8612 **	0.9527 **	0.9180 **
Asymmetry factor	−0.3853 **	−0.3183 **	−0.0971	−0.1565 **	−0.3548 **	0.2886 **
Speculation index (1)	−0.0161	−0.0071	−0.0374	0.1780 **	0.0108 *	0.0432
Month (2)	0.0461	0.1376	0.0631	0.1095	0.0234 **	−0.0477
1 × 2	0.0445	−0.0832	0.0825	0.1538	−0.0557 **	0.1775
Llik:	−17,487	−15,239	−16,362	−18,605	−11,911	−19,189
BIC:	35,057	30,559	32,807	37,292	23,905	38,460
AIC:	34,993	30,495	32,743	37,228	23,841	38,396
HQC:	35,014	30,517	32,765	37,250	23,862	38,417

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.1 are flagged with one asterisk (*), and those with p-value less than 0.05 are flagged with two asterisks (**).

). The mean equation’s parameter estimates are similar to the prior TGARCH model. All constants are statistically significant exactly as in the previous model, although the parameter for autoregression is statistically significant in soybean and gold but not in oil. Both the residual and volatility parameters are statistically significant in this model, as they were in the previous one. The constant in the variance equation is also statistically significant in all cases. Except for corn, the asymmetry factor is statistically significant in all cases. Except for oil, all of the asymmetry parameter values are negative. This means that when returns go down, volatility goes down more than when returns go up. Only in the case of oats and gold futures does short-term speculation have a statistically significant and positive effect. Gold has a statistically significant seasonal multiplied effect, but it reduces return volatility rather than increases it because the parameter has a negative sign (estimated to be −0.0557). Note that gold is the only commodity that has a statistically significant monthly effect in the model (estimated to be 0.0234). The smallest information criterion is for gold, whereas the largest is for oil. With the exception of oil, this model’s information criteria are the lowest when compared to the prior TGARCH model, indicating that it more accurately captures the returns from futures contracts.

The GARCH modeling results are then examined using only the post-2020 data. To begin, we look at the results of the classic GARCH model to see if there are any differences (

Table 8. Parameter estimates of the GARCH model using Approach II for post-2020 data.

Indicator	Wheat	Soybean	Corn	Oats	Gold	Oil
Mean equation:
Constant	0.0298	0.0757	0.0886	0.1393 *	0.0284	0.1984 **
Return	−0.0779 *	0.0492	0.0479	0.1261 **	−0.0304	0.0154
Variance equation:
Constant	0.1616	0.0280	0.1396	0.3158	−0.1019	0.1939
Residual	0.0285	0.0955 *	0.0740 **	−0.0501 *	0.0453	0.2620 **
Volatility	0.8629 **	0.8704 **	0.8858 **	0.8669 **	0.8287 **	0.7534 **
Speculation index (1)	0.4531	0.1333	−0.1361	1.2090	0.5561 *	0.2781
Month (2)	4.8277 **	−0.0868	−0.6826	0.3038	3.4175 **	8.9741 **
1 × 2	−18.3034 **	0.4176	3.0159	2.1948	−5.9323 **	−24.3412 **
Llik:	−838	−713	−832	−834	−642	−1076
BIC:	1725	1474	1712	1716	1332	2201
AIC:	1693	1442	1679	1684	1300	2169
HQC:	1705	1454	1692	1697	1312	2181

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.1 are flagged with one asterisk (*), and those with a p-value less than 0.05 are flagged with two asterisks (**).

). The mean equation’s parameter estimates are comparable to those in the earlier model. However, in this instance, the return autoregression parameter for wheat becomes statistically significant (estimated to be −0.0779) whereas the parameter values for corn and soybean are not statistically significant. Volatility is statistically significant for all six commodities at a p-value of 0.05 or less. However, in wheat and gold futures, the residual is statistically insignificant. The constant is statistically insignificant in any case. Only in the case of gold does short-term speculation have a statistically significant and positive effect (estimated to be 0.5561). The month has a statistically significant effect on wheat (estimated to be 4.8277), gold (estimated to be 3.4175), and oil (estimated to be 8.9741). The multiplied effect is statistically significant for all these three commodities as well and has a negative sign, showing that speculation reduces volatility during more volatile time periods. The smallest information criterion is for gold, whereas the largest is for oil.

The TGARCH model is then used to investigate the same time period with the asymmetry factor (

Table 9. Parameter estimates of the TGARCH model using Approach II for post-2020 data.

Indicator	Wheat	Soybean	Corn	Oats	Gold	Oil
Mean equation:
Constant	0.0410	0.0833	0.0961	0.1454 *	0.0322	−0.0278
Return	−0.0864 *	0.0334	0.0460	0.1264 **	−0.0321	0.0722
Variance equation:
Constant	1.3589	0.0599	0.0396	0.4717 *	0.0035	0.3963
Residual	0.0011	0.1129 **	0.0686 **	−0.0593	0.0468	0.1697 **
Volatility	0.5289	0.8728 **	0.9345 **	0.8419 **	0.8553 **	0.8342 **
Asymmetry factor	−52.3972	0.0002	−0.3199	0.0745	−0.1281	0.9000 **
Speculation index (1)	−0.1311	0.0497	−0.0049	0.6889	0.3118 **	0.8470
Month (2)	2.8169 *	−0.2448	−0.2379	0.2363	1.9869 **	8.3801 **
1 × 2	−10.3698 **	1.0337	1.1387	0.5469	−3.4575 **	−21.0734 **
Llik:	−838	−708	−831	−834	−639	−1062
BIC:	1730	1471	1717	1724	1333	2178
AIC:	1693	1434	1680	1687	1297	2141
HQC:	1708	1449	1694	1701	1311	2156

Source: author’s calculations based on CBOT [90], NYMEX [91], and COMEX [92] data, 2021. Notes: Estimates with a p-value less than 0.1 are flagged with one asterisk (*), and those with a p-value less than 0.05 are flagged with two asterisks (**).

). The mean equation’s parameter estimates are statistically significant only for oats and wheat, but for wheat only the return parameter is statistically significant (estimated to be −0.0864). In the variance equation, the volatility parameter is statistically significant in all cases but wheat. Except for wheat and oats, the residual parameter is statistically significant. The constant is statistically significant for the oats futures. Only in the case of oil is the asymmetry factor statistically significant and positive, indicating that negative returns are followed by increased volatility (estimated to be 0.9000). Only in the case of gold futures does short-term speculation have a statistically significant and positive effect (estimated to be 0.3118). Wheat, gold, and oil all have a statistically significant but negative effect on volatility during the most volatile months, indicating that increased short-term speculation reduces volatility during the most volatile months. The smallest information criterion is for gold, whereas the largest is for oil. If comparing this to the previous GARCH model for post-2020 data, information criteria are smaller in the GARCH model for soybean, gold, and oil futures.

We conclude that realized futures returns can be effectively modeled using GARCH techniques using similar methods to those of other authors [16,19,93]. In most cases, time series are stationary, with residual and volatility effects present. Furthermore, we estimate cases of negative asymmetry effects, which means that an increase in returns is more likely to be accompanied by an increase in volatility than vice versa. Short-term speculation has the greatest impact on volatility in oats and gold futures. Seasonality has a greater impact on agricultural commodities than on gold or oil. However, season and speculation effects do not affect returns for agricultural commodities except for wheat when using post-2020 data. Finally, for the majority of commodities, we are unable to pinpoint the disruptive impact of financial speculation on underlying futures price volatility. In accordance with the results of our research, financial speculation more often had no impact than increased volatility or had a neutral effect. Both subsamples produce comparable results, yet there are cases when speculation does have a negative effect on return volatility—most notably in wheat, gold, and oil. This has important practical implications, as speculation that co-moved with increased prices in many commodity markets is not necessarily responsible for price spikes and volatility of these products. Therefore, commodity futures markets can continue to grow and develop regardless of speculative activities inside them and provide financial tools to manage price risks for agricultural producers and marketers, thus resulting in a more sustainable supply of agricultural products.

4. Discussion

4.1. Relevance of Current Findings in Light of Earlier Studies

This study, like those of other authors, focuses on the volatility of futures contract returns and the variables that influence them. This research employed an autoregressive conditional heteroskedasticity model with the speculation index (TV/OI) in the variance equation to estimate the influence of speculation on agricultural and other futures returns. The research provides several major findings.

Speculation in most commodities used in this study, except for gold and oil, does not drive return volatility. However, there is some evidence when analyzing the crude oil market that speculation does have a positive impact when using full sample data and applying the traditional GARCH model. Similarly to other authors that use various measures for speculation activities, prices and returns from futures contracts are more often unaffected by speculation activities [94,95]. However, some studies observe alternative relationships. For example, Wellenreuther and Voelzke [6] found that when TV/OI is used as a measure of speculation, returns seem to increase speculation in Chinese markets (Dalian Commodity Exchange). Bohl et al. [37] also used TV/OI as a measure for speculation, and it increased volatility in soybean and corn markets, but they also analyzed Chinese markets. In addition, according to research done by Czudaj [16], open interest and trade volume from the prior period influence how volatile agricultural futures markets are (trade volume and open interest are used separately). Algieri and Leccadito [22], who use the scalping index, found that speculation explains volatility in less liquid markets, such as pork bellies, orange juice, feeder cattle, as well as sugar, ethanol, and natural gas, but not when analyzing panel data for commodity groups. The statistically significant but negative effect of speculation on returns for crude oil is also noticed in other authors’ research [96]. Furthermore, when analyzing and applying the GARCH model in non-US commodity markets, a statistically significant effect of speculation on oil and gold return volatility is discovered [14]. According to Guo et al. [97], speculation has a detrimental impact on oil prices. According to research done by Borin and Di Nino [98], return volatility from sugar, wheat, and cotton markets is positively affected by long swap fund positions. When analyzing data for 1986–2010, speculation in the oats market had a strong effect, whereas others (crude oil, soybeans, and corn) had small but positive effects [15]. Similarly, older studies using Granger causality tests discovered that wheat and corn trade volumes explain price volatility [27].

Another important finding in this paper is that speculation, in some cases, is found to stabilize the return volatility of certain commodities. Other authors also found these effects in the corn market but did not use the TV/OI index as a speculation indicator [94]. According to Bohl and Sulewski [19], the share of speculation positions stabilizes prices in agricultural commodities, especially corn markets. Kim [93] used the GARCH method to look at data from 1992 to 2012 and found that speculation had a similar effect on the markets for wheat, soybeans, corn, and especially crude oil. When analyzing data for 1986–2010, Manera et al. [15] observed that returns from wheat futures were lowered by the scalping index. According to the research, long-short traders help to improve price discovery by lowering overall market volatility and cross correlation with equities [99]. According to Xiao et al. [100], the crude oil futures market’s speculative behavior helps protect against shocks caused by macroeconomic uncertainty.

The next important aspect when analyzing this and other authors’ research is the selection of time period or dividing time samples into subsamples, such as before and during the economic turmoil. On the one hand, some authors, such as Hachula and Rieth [34], argue that demand shocks from index investors have a very minor impact on the dynamics of futures prices, both generally and during the boom–bust cycles of 2007–2008 and 2011–2012. On the other hand, according to Haase and Huss [101], speculation lowers price volatility, especially in difficult economic circumstances (for example, the wheat market), but they used Working’s T index for speculation activity. This is similar to our findings for the wheat market using the post-2020 time period where speculation during the more volatile months is found to reduce return volatility. Algieri [17] points out that wheat price volatility has been caused by excessive speculation in certain time periods, although not all commodities are affected by the same linkages. Bohl and Stephan [102] found a positive effect on return volatility when using GARCH modeling in US corn, wheat, and soybean markets as well as crude oil when using trade volume until 2002 but not after, when major events considering financialization occurred in the commodity markets.

Our research also analyses seasonality’s impact on return volatility, which is highlighted by other authors and is common in agricultural markets [103]. Silveira et al. [78] analyzed returns from corn and soybean futures contracts and concluded that return volatility increased during the second quarter of the year for corn and soybeans but before 2004. According to Gupta and Rajib [104], when analyzing Indian markets, the volatility of the futures contract depends on the time of the year. Karali and Ramirez [8] used the GARCH model to predict the volatility of energy futures returns and observed that October is the most volatile month for crude oil when analyzing data from 1994 to 2011. According to Karali and Thurman [77], when analyzing 1986–2007 data, systematic seasonal elements with increased volatility before harvest seasons are observable within corn, soybeans, wheat, and oats futures contracts, and show slightly different patterns in the GARCH model. According to Peterson and Tomek [86], the volatility of corn prices traded in the US is highest in May through to August and lowest in November. Smith [87] also notes that the US corn market is characterized by increased price volatility in September.

The next important observation in this paper is that an asymmetry between returns and volatility is visible in most commodities, except for corn. However, the post-2020 data only show this effect for oil. In addition, the parameters’ values, except for oil, are negative, showing that negative news reduces further volatility rather than increases it. Other authors who applied asymmetric GARCH models found similar results and that asymmetry does exist in the US grain markets [15,96]. According to a study done by Silveira et al. [78] analyzing data for 1959–2014, the TGARCH asymmetry parameter for corn was negative until 2004 but positive after. For soybeans, it was negative until 2004 but insignificant later. Baur and Dimpfl [105] claim that when they looked at data from 1990 to 2017, this parameter was significant and positive for crude oil and corn after 2004, but it was significant and negative for soybeans, wheat, and gold before 2004, yet oats futures were not used in the study. According to Garefalakis [106], oil price volatility rises more after a significant negative shock than it does after a significant positive shock.

Another important finding is that when analyzing the post-2020 timeframe, season-weighted speculation reduces volatility in oil and wheat, but there are mixed effects in the gold market. As mentioned above, the oil market also shows asymmetry in the post-2020 timeframe. This may be the result of panic in the markets. For example, according to Salisu et al. [38], there is a correlation between the global fear index and commodity price returns, with the latter increasing as COVID-19-related anxiety increases. According to Ahmed and Sarkodie [39], in both low-volatility and high-volatility regimes, the empirical data demonstrate that the returns on commodities, such as oil, natural gas, corn, soybeans, silver, gold, copper, and steel, adapt to shocks in COVID-19 results and economic policy uncertainty to variable degrees. In addition, Borgards et al. [107] demonstrate that, for the commodity futures under consideration, the overreaction hypothesis is true, and overreactions occur more often and with greater amplitude. However, soft commodities exhibit far fewer overreactions than precious metals and energy commodities. Because there were more negative than positive overreactions to the COVID-19 outbreak, crude oil futures in particular display a distinct overreaction tendency from other commodities [107]. In contrast to the pre-COVID-19 period, the return spillover was more pronounced during the COVID-19 crisis [40]. According to Wen et al. [108], the impact of the stock market on the commodities market has dramatically increased during the COVID-19 epidemic.

4.2. Model Improvement and Future Research Recommendations

This leads to the conclusion that short-term speculation has a mixed impact on selected commodity futures and that the destabilizing effect when analyzing agricultural commodities is unique to the oats market. If both short-term speculation and return volatility are influenced by changes in energy costs, then future studies should also stress their co-movement with energy prices.

Further investigation may be as follows: to extend the investigation to other commodity exchanges; to supplement the models with new variables describing speculative activity; to apply more models that allow more accurate modeling of the impact on return variability (EGARCH, TGARCH, etc.); to establish causal relationships (Granger causality test); and to assess the effect of speculation on the co-movements between the volatility of commodity returns.

The next important recommendation for future research is that energy price movements can be analyzed using their impact on agricultural commodities. Besides the fact that biofuel is used to produce ethanol, corn prices have grown increasingly tightly with ethanol and its impact on corn prices can be further extended [94]. This then can be expanded on how speculation affects correlation among different commodity markets. For example, according to a study by Fan et al. [99], the cross-speculative pressure is still minimal and greater speculation does not lead to the correlation of commodities that at first seem to be unconnected, but the post-COVID time period is not yet used in the research. Additional factors can be added to the model, considering speculative pressure as well as the global fear index used by others [38], such as the macroeconomic uncertainty index [100]. Post-macroeconomic uncertainty variables can be analyzed as well to find better explanations for increased return volatility and what factors do cause increased speculation activity.

Research on asymmetric relationships can be expanded by using more commodities and more time sub-samples similar to studies by other authors [78]. According to Lawson [1], the effect of speculative activity on product prices varies depending on the commodity (rice and wheat prices respond to speculative activity less than corn and soybean prices) and the variable used to measure speculative activity. For instance, Zivkov [76] claims that soybean futures are the best instruments for diversification because their values are least dependent on oil prices and their inherent volatility and spikes. The investigation can be further expanded when focusing on seasonal changes using more subsamples. Additional dummy variables can be added to the variance equation besides seasonality, such as economic shocks, periods of increased market volatility, and others. Dramatic changes took place during the Russia–Ukraine war. Contemporary research emphasizes the impact of these factors on how speculation activity reacts during spikes in energy prices. Therefore, the Russian–Ukrainian war’s effects on the commodity market can also be assessed. Alam et al. [109] claim that speculation-driven shocks to oil demand are statistically meaningful primarily under regimes of high volatility. During the economic crisis, agricultural futures are more susceptible to long-lasting price bubbles with significant price changes (soybeans and cotton suffered price bubbles from 2007 to 2008 and from 2010 to 2011, respectively) [110].

Another important observation is that other authors compared long-run and short-run indices of financial speculation [22]. Yet this approach requires enough weekly data for a long-run speculation index to better measure the pandemic and post-pandemic eras. As new data on commodity markets become available, it is necessary to conduct research on long-run speculation indicators such as Working T’s index for excessive speculation, the ESV index, long and short commercial positions, and others employed by other authors in their research on commodity markets [34,101]. It is also important to assess how long-run speculation reacts to seasonality. In this way, it is possible to complement and extend the existing model, which shows how short-run speculation affects the volatility of returns in those months when these markets are characterized by the highest volatility of returns. Some authors observe that long-run speculation reduces return volatility, but this needs further analysis [111]. Thus, an increase in index positions may reduce this excessive volatility. According to Manera et al. [15], short-term speculation raises volatility, whereas long-term speculation lowers it. Therefore, the impact of speculation on prices, according to Lawson et al. [1], is diverse. The influence of speculation on the price of food grains, however, is not uniform and varies depending on the definition of speculation and the commodity.

Research can also emphasize trading in other markets that are growing in size, such as European commodity markets. The authors emphasize that different types of impact were found in Chinese markets [6,37]. For example, the oats market is smaller and less liquid, so it is worth analyzing if speculation is more destabilizing in markets such as these or, for example, cattle or dairy markets, as in studies by others [35,70,112]. Some observe [65] that the sugar market is also different and unique, so it should get more attention in future analyses. Soft and livestock commodities have demonstrated the important advantages of diversification for investors in the energy sector. In addition, hedgers’ availability of liquidity enables speculators to take positions that influence the returns of global benchmark prices in soft commodities [113].

Non-causality tests such as the Granger test can also be used to learn more about the connections between returns and speculation, especially in the oats market, which other authors, besides some exceptions [64,114], did not pay enough attention to. To get a better description of these interactions, however, one may employ the GARCH-M model for projecting risk premiums [96]. As mentioned above, different time periods show conflicting results; therefore, a rolling Granger test can be employed like authors [74]; DCC GARCH models [115]; and the Markov-regime switching model [109]. An analysis of panel data can be employed as well [22]. More technically sophisticated and innovative methods, such as neural networks (NN), can be used, similar to other studies using time series data [116] for a better assessment of the long-short-term memory in the return time series. These methods can also be used to assess seasonal effects (seasonal fluctuations) to see if these additional variables, including months or speculation, bring higher accuracy to forecasting [117]. It is also important to look for structural breaks in the time series of agricultural commodities [118,119]. The non-causality test may be used in more sophisticated ways, like in the Diks and Panchenko’s (2006) [120] test, by demonstrating that these interactions are nonlinear and by assessing if the residual errors of the models are connected. The research can then be expanded on how financialization processes and other external factors drive the dynamics of commodity futures markets. A noteworthy finding is that the impact of seasonality on return volatility was less pronounced in the post-2003 period than it had been in earlier periods. This was due to the development of commodity exchanges around the world and the intensification of interconnections across financial markets. For long-term memory effects on return volatility, more complex GARCHs like E-GARCH or AP-GARCH can be used, similar to studies by Czudaj [16]. In an environment where commodities used to be more financialized, others find that basic market impacts prevail [121]. However, there is some indication that speculators had a greater role during the time most closely related to financialization [122]. Another problem to investigate in future research, the consequences of which already exist and can be predicted, is poverty, political instability, and famine in individual regions due to price spikes in commodity markets.

4.3. Practical Guidelines

Futures contracts provide an opportunity to hedge against price risks for producers and marketers who face abnormal market risks [123]. Another crucial factor to take into account is the possibility that commodity traders who are more susceptible to price fluctuations may call for more regulations on speculative behavior in the commodity markets [124]. This creates the need for more restrictions on commodity exchanges. Problematic questions remain about what can be done to ensure sustainable development and food security. Interventions in the commodity exchanges or more market freedom? How important is it for the government to have food reserves? What can be done to make sure that market signals have a balancing effect on demand and supply rather than a shocking effect?

For this reason, it is important to differentiate between active and passive approaches on how to regulate commodity exchanges around the world. Marginal account requirements, transaction fees, and position restrictions in US and EU product markets are only some of the active mechanisms to be used. Tougher reporting requirements, more openness in product markets, and stricter regulation of over-the-counter transactions are all examples of passive measures. Active interventions, empirical studies show, fail to achieve their declared goals and instead cause greater price volatility, and many authors argue that passive measures can be more effective and more based on research [19,20].

Other researchers focus on trade restrictions and their impact on price and return volatility. At this stage of the development of the product market, according to those academics who have documented particular instances in which speculative considerations have destabilized prices, bans or other restrictions would be detrimental to market growth [14]. The capacity of commercial market participants to handle price risk is constrained by the capital constraints of non-commercial market players, which has an impact on futures and product pricing [125]. Others contend that position limitations will cause hedging demands, price volatility, liquidity, and the risk premium to be more skewed [99].

On the other hand, it is crucial to provide market participants and policymakers with reliable data on the various types of traders and their significance for the efficient functioning of the market and pricing system [126]. Exchanges for commodities should provide more frequent data, more detailed information on trading processes, and signs of market concentration, and they should also divide market participants into commercial and non-commercial participants. Reduced product market uncertainty and price volatility might be achieved by more financial market openness and a more transparent legal framework [22,50]. This is especially true for other countries’ commodity exchanges, where few academics have used these markets for empirical research because of a scarcity of data and the difficulty of identifying all indications of speculative activity. Everyone involved in the market and the relevant authorities should be made aware of the findings. Others also find that long–short speculators ethically invest in agricultural commodities, representing the planet’s most vital food supplies, and they tend to reject the idea that commodity prices are being artificially manipulated [29].

A well-developed commodities market with liquid futures contracts might be of assistance in mitigating the effects of external shocks like the pandemic-induced economic turmoil. Establishment or expansion of already existing commodity markets provides online trading in a variety of commodities, which helps investors broaden their investment horizons, benefit from portfolio diversification, and lower their risk of losing money [12]. A strong commodities futures market, for instance, allows farmers to protect themselves against the kinds of price fluctuations and income uncertainties that are part of the course in agriculture, particularly during epidemic years. Second, the supply chains of agricultural goods, which are vulnerable to health crises like these and contribute to food security, are protected thanks to the commodity futures markets, which benefit not only farmers but also agricultural producers who hold long positions to hedge against rising prices. Futures contracts, on the other hand, may be employed during times of uncertainty such as increasing energy costs, meaning that their usage is still profitable after the pandemic has passed. Besides the oats market, others provide evidence that soft commodities look to vary from their inherent values and propose that governments may think about restricting speculation or hedging to deflate the bubble but use soft commodities in their research [18]. According to the results of our research, we support the position of other authors and believe that a general (active) regulation of all commodities is not the best course of action given the many beneficial consequences of financialization [18].

The results of this study are important in the context of sustainable development. Sustainable development involves the management of various risks. Derivatives used by producers and processors of agricultural products help manage market and price risks, so it is important that the exchanges of these instruments function harmoniously and sufficiently. Therefore, financial instruments such as commodity futures contracts can help maintain the viability of agricultural supply chains, as agricultural producers are given the means to manage and be less exposed to price and market risks. This helps to protect the most vulnerable parts of agricultural production chains, which are more sensitive to risks. In addition, liquid and well-functioning financial markets allow prices to adjust more quickly to new events and thus avoid excessive price shocks.

5. Conclusions

In our research, we evaluate the relationships between financial speculation and returns from commodity futures in agriculture and other commodity markets to provide implications for the sustainable development of agricultural markets. We use daily returns on wheat, soybeans, and corn futures from the Chicago Mercantile Exchange. The results demonstrate stationary time series and volatility clustering. Therefore, further modeling of exogenous variables using ARCH techniques is available. More specifically, we adopt generalized autoregressive conditional heteroskedasticity (GARCH) models, including the asymmetric TGARCH approach, and focus on the impact financial speculation has on return volatility. We use the short-run speculation index (TV/OI) to measure speculation. We also include seasonality into our models, assuming that springtime agricultural futures prices are more volatile than other seasons and that negative returns have an adverse influence on future return volatility.

The results of our preliminary and proposed model analysis show three important findings. First of all, among all agricultural commodities, only the results from the oats market show significant signs of speculation destabilizing prices. Even though some evidence is found in the gold market, the crude oil market shows mixed results. Second, all commodities except soybeans show an asymmetric effect from the previous return to conditional volatility. However, parameter values for all commodities except for crude oil are negative. This shows that negative returns are followed by less volatility than vice versa. Third, returns from wheat futures contracts are reduced by the short-run speculation during the more volatile month of May. Similar results are observed in the gold market during the month of January and the crude oil market during the month of November.

The results of our research might have significant effects on legislation and public policy relating to commodities exchanges. The authorities of the futures and commodities markets are concerned about financial speculation, which is why they suggest imposing restrictions on regular operations, margin accounts, and other matters. Our research, along with that of other authors, suggests that the influence of financial speculation on the level of prices and the volatility of returns in agricultural markets is, at best, debatable, and in some instances, may even have the opposite effect; for example, an increase in short-horizon trading may be followed by prices and returns that are less volatile. Well functioning and liquid commodity exchange markets provide hedging opportunities for commercial market participants. Future research should focus more on other measurements for speculation, as well as better analysis of economic turbulent time periods.

The results of the study show that financial markets and commodity exchanges, despite their growth and expansion, remain little affected by speculators present in these markets. In order to secure the supply of agricultural products, it is important to reduce risks in this sector, which increase production costs, and to create conditions for attracting additional investments. This highlights the role and importance of commodity markets, also emphasizing the further development of these markets. Financial markets help distribute risks, and agricultural producers and traders can be less exposed to price risk, which increases with various economic shocks caused by health, political, and other crises. Financial instruments provided by the financial markets can ultimately lead to lower production costs, a more sustainable agriculture sector, and create conditions for greater production of various commodities. Therefore, further attention should be paid to the sustainable development of financial markets.

Model improvement and future research recommendations. The model applied in the research can be improved adding new variables describing speculative activity (short-run and long-run speculation) to the models, applying more sophisticated GARCH models that allow for more accurate modeling of the impact on return variability (EGARCH, APGARCH, etc.), establishing causal relationships (the Granger causality test), and evaluating the impact of speculation on the co-movements between the volatility of commodity returns. By utilizing more commodities and time sub-samples, research on asymmetric interactions can be expanded. Research can also highlight trading in other expanding markets, such as European commodities markets.