Financialisation of Commodity Markets: Evidence from India

Abstract

The known circumstances that favour financialisation of commodity markets which result in unidirectional co-movement of equity and commodity indices are either weak or non-existent in India. Yet, after 2015, there has been a greater correlation between equity and commodity markets even when decoupling is observed in global markets. Results from the rolling regression attest to the shift in response of commodity and equity indices to wholesale price inflation (WPI) and call rate after 2015, indicating that post 2015 co-movement could have been a result of inflation targeting regime. The linear regression as well as the Granger causality analysis based on vector autoregression (VAR) framework, which accounts for simultaneity, confirms that commodity markets are moving on its own supply-demand factors. The rolling regression also brings to light the disciplining effect of regulatory scrutiny and audit trail in the Indian commodity market around July 2013, when National Spot Exchange Ltd. (NSEL) payment crisis and commodity transaction taxes (CTT) occurred.

JEL Classifications: G180, G28, C580, G1, G100. G130

Keywords

Inflation targeting Financialisation Transaction taxes NSEL CTT Commodity markets

1. Introduction

In general, commodity prices are believed to move in a direction opposite to that of equity prices. Hence, exposure to commodities is considered a natural hedge against equity price risks (Gorton & Rouwenhorst, 2006). For this reason, the commodity derivatives segment has emerged as an alternate asset class for their risk diversification for financial players. The hypothesis that the flows of financial commodity investors impact commodity markets is referred to as the “financialization of commodity markets” (Henderson et al., 2015). Certain studies (Awasthi et al., 2020; Kumar & Somnath, 2014; Nissanke, 2012; Sen, 2014; United Nations Conference on Trade and Development [UNCTAD], 2012) have shown that with the entry of financial players such as index investors and mutual funds into the commodity derivatives market, the traditional negative correlation between commodities and other financial assets breaks down and both start moving in the same direction. This could happen mainly because of the following factors:

The trading strategies used by financial players depend more on financial or general economy-related information and less on specific commodity market fundamentals. Unlike equity markets where companies are mandated to disclose their price impacting decisions first or simultaneously with stock exchanges, in the commodities segment, information is fragmented across various local spot markets and may never come forth or be available to all the players at the same time. Commodity research is still at its infancy. Hence, the entry of financial players causes commodity markets to diverge from commodity-specific supply-demand fundamentals and markets begin showing herd behaviour, following the logic of financial asset markets (Mayer, 2012; UNCTAD, 2012).

At times, decisions on many commodities for which specific information is not available would be taken based on the information readily available on a commodity like crude oil, which affects all commodity transports. Commodities like oil and gold are also influenced by financial interests, and herd behaviour causes the rest of the commodity markets to follow suit.

Concurrent with the rapid growth of index investment in commodity markets, prices of non-energy commodities have become increasingly correlated with crude oil. Since index traders buy as a basket, if a purchase of an index is made on account of fundamentals in the oil market or in metals, additional purchases are made in other commodities constituting the index, thereby driving up all commodities together (Tang & Xiong, 2012; UNCTAD, 2013). If real hedgers’ participation is below some threshold, then price formation would be largely based on the logic of financial investors (Plantier, 2013).

Further, institutional investors are instrumental in bringing in programmed trading or algorithms and trend-based automated trading strategies, which could further increase the volatility and amplitude of business cycles. This eventually leads to sharper commodity boom-bust cycles (Bicchetti & Maystre, 2012).

This article seeks evidence of financialisation of the commodity derivatives markets in India, while examining the global evidence in this regard.

2. Existing Global Evidence

Many scholars have attributed the rise in global commodity spot prices during 2002–2008 to financialisation trends. For instance, the UNCTAD stated that the liquidity benefit from financial players’ participation is overshadowed by the increase in hedging costs to the market participants due to increased volatility in commodity prices. The ‘risk of speculative bubbles and prolonged deviations from fundamental values rises accordingly, with a potential for distorting economic activities and causing financial fragilities through spill over risks’ (UNCTAD, 2012). However, some scholars have challenged the financialisation of commodity markets citing a lack of robustness of the evidence, particularly of the negative influence of index investments on commodity markets (Hamilton & Wu, 2015; Irwin & Sanders, 2011). After examining the current state of price formation in commodity markets, Valiante and Egenhofer (2013) confirm that demand and supply fundamentals remain solid drivers of futures price formation across all the commodities markets covered in their report, though metal and energy futures are heavily influenced by financial players. Their report suggests that, so far, the role of non-commercial participants in commodities’ markets has been generally benign, and the growth of index investments has not caused distortions in price formation.

Studying trader positions in 17 US commodity futures, Büyükşahin and Robe (2014) show that the correlation between the rates of return on investible commodity and equity indices rises, amid greater participation generally by speculators, especially hedge funds, and particularly hedge funds that hold positions in both equity and commodity futures markets. They did not find any such relationship for commodity swap dealers, including index traders. According to them, the characteristics of traders help predict the joint distribution of commodity and equity returns. Rossi (2012) investigated the predictive power of commodity prices using equity prices and found that the coefficient was mainly positive in the period around the early 2000s, possibly reflecting the high demand pressures on expectations of increasing growth-induced future returns; however, towards the end of the sample, around the start of the financial crisis, the coefficient becomes negative, possibly reflecting higher risk aversion.

3. Is There a Co-movement of Global Equity and Commodity Indices?

Most of the studies that confirmed the co-movement of equity and commodity markets study the period before 2012. The latest set of data (Figure 1) points to decoupling of the two markets at the global level, from around 2012 onwards. To understand the co-movement of both sectors, the most widely used commodity indices—the Bloomberg Commodity Index (BCOM) and an equity index, MSCI World¹—are matched for the period 1980–2019 (Figure 2), along with their 252-day moving average (average trading days in a year). A correlation matrix for the various sub-periods is given in Table 1.

Figure 1

Source: S&P Indices,² https://www.spglobal.com/spdji/en/indices/equity/sp-global-bmi/#overview

It can be seen that over the past four decades, coupling of the markets and synchronised movements were visible mainly over the decade 2002–2012, holding firm through troughs and crests. Thereafter, the markets have decoupled again. The coupling starts immediately after the bursting of the dot com bubble in 2000 and through its recovery phase to the next full-fledged global financial crisis in 2008 and the immediate V-shaped recovery thereafter. It can be seen that except for the said period, the correlation coefficients were either negative or at very low levels. This has also been confirmed by Bianchi et al. (2020), who using quantile regression, have pointed to aggressive financialisation during 2004–2013 and de-financialisation among non-energy products during 2014–2017.

Figure 2

Source: MSCI and Bloomberg (daily data).

Table 1.

Correlation between the MSCI Index and BCOM Index, 1980–2019

Period	Correlation Coefficient
1980–2019 (long period average)	0.32
1980–1990	−0.42
1991–2000	0.04
2001–2005	0.16
2006–2012	0.80
2012–2019	−0.61

Source: The author.

While this correlation is confirmed in many studies, it is not yet conclusive as to what drives or causes it. Considering that 2004–2007 was a period of robust global growth at nearly 4%, it could very well be the case that strong commodity (especially agricultural) prices (spot and futures) may be a concomitant result of the economic growth, which normally drives equity indices up on account of expected positive future cash flows. Commodity indices could have risen in anticipation of greater consumption of commodities.

UNCTAD (2012)and many others attribute the co-movements to the activity in Over-The-Counter (OTC) markets, which saw a substantial build-up in global OTC commodity derivatives through 2005–2008, followed by a sharp decline until 2010 (Figure 3).

Figure 3

Source: Bank of International Settlements (half-yearly data till December 2018).

While one is inclined to rely on this explanation, it is pertinent to note that global commodity spot prices had, on average, started moving up since 2002 (Figure 4), whereas OTC derivative activity started building up only around 2005. The steady rise in spot prices could have fuelled hedging using derivatives. It is often reported by market participants in India that a price rise, more than a price fall, triggers an increase in activity in the derivative segment. Commodity derivatives constituted around 11%–12% of the gross market value of all the OTC derivatives by around December 2007 to June 2008, compared to around 1%–2% during normal periods. However, in June 2008, around 95% of such commodity derivatives were structured on commodities other than gold or precious metals, the normally considered safe haven assets that attract financial investors.³ Further sub-categorisation into what percentage comprises metals, energy or agricultural derivatives, is not available. Nor is there is any granular data as to what percentage of the OTC turnover was from institutional/financial players and real hedgers.

Figure 4

Source: Compiled from UNCTAD statistics.

While admitting that OTC contracts may have played some role, it is pertinent to note that the period 2002–2012 represents the first period of synchronised global cycles, wherein globally, countries seem to be moving together into a recession or a boom, aided by capital outflows and inflows that guide financial market benchmarks. Before this period, a crisis used to be limited to certain countries or regions (e.g., the Asian financial crisis of 1997, Russian crisis of 1998, Argentine crisis 1999–2002, etc.). Capital flows, trade flows and information flows could have played a major role in this synchronisation of business cycles across the globe. Figure 2 shows that the shift from peaks and troughs as identified through the 252-day moving averages happens first in equity markets, followed by commodity markets during this period. It could be that, in the chaos prevailing during a financial crisis, expectations about potential global growth that drive equity markets in a particular direction would have also moved commodity markets in the same direction, in anticipation of greater/lower commodity consumption. Could it be then the case that the pairing up of equity and commodity markets in the same direction is an aberration created by a lack of information on where the world is heading, particularly in a bubble phase or during sudden crashes? In other words, ceteris parabis, a positive correlation particularly between the non-agricultural segment (as the agriculture segment is found to be often correlated with the equity market) and equity market could be a sign of an impending bubble and/or its bursting. Devoid of such events, an acceleration phase of the economy shows a strong negative correlation of commodity markets with equity markets. A recent study (Awasthi et al., 2020) has also observed that co-movements between equity markets and their sample commodities were more pronounced during a crisis period. According to Adams and Glück (2015), if uncertainty among investors causes a commodity to trade at a high volatility, the commodity shows an increased exposure to risk spillovers from stock market shocks.

4. Is There Financialisation in India?

As far as India is concerned, the enabling conditions for financialisation of commodity markets were either weak or non-existent till 2019, for which the analysis has been carried out. Institutional players were not allowed to participate in the market even as late as June 2017 when category III Alternate Investment Funds (AIF or hedge funds) were permitted, and at the end of March 2019, only one hedge fund was registered. Commodity funds floated by mutual funds were non-existent as mutual funds and portfolio managers were permitted into the commodity segment only in May 2019. Going ahead, mutual funds as well as hedge funds are not allowed to participate in any of the sensitive commodities and are effectively limited to trading in gold. Foreign Portfolio Investors (FPIs) are still not allowed to participate. However, foreign investors have been allowed, but only as recently as October 2018 and limited to pure hedging activities, and none had opened an account till March 2019. Commodity index trading commenced only around June 2020, although exchanges had floated a few indices since 2006–2007. The quantum of OTC contracts in commodities is not known.⁴ Algo trading is permitted in both the equity and commodity markets. The only potential domestic channel of spillover effects from the equity segment in India comes through the proprietary trading desks of brokers who are active in both the equity and commodity segments. Nearly 50% of the trades are reportedly emanating from non-value chain participants, and hence, are “financial” in nature.⁵ Another potential reason could be that the metal and energy contracts traded in India are mirror contracts, as they are settled on the prices discovered in the London or New York exchanges, thereby reflecting global commodity movements, which were co-moving with equity markets.

Figure 5.

Source: NSE, NCDEX and MCX.

For a deeper examination, the NIFTY—the major equity index consisting of India’s top 50 stocks —is compared in Figure 5 with the major commodity derivative indices: Dhaanya, a pure agri index consisting of futures prices of ten agri commodities traded at NCDEX⁶; and Comdex, a composite index with 40:40:20 weightage, respectively, for energy, metals and agri futures prices at the MCX. Since 2012, the MCX has been creating an actual return-based index for commodities—iComdex—which is also used for examination. In iComdex, the futures returns based on the previous day’s futures prices are considered, rather than the futures prices of a base-year period, as is the case in other indices (NIFTY, Dhaanya or Comdex). iComdex also has some single commodity indices, of which crude oil and gold are considered for examination. Further, the relevant correlation coefficients for the various periods of importance to commodity markets are given in Table 2.

Table 2.

Correlation of Commodity and Equity Market Indices, 2006–2019 (based on market specific issues)

Correlation Matrix
Period	Features of the period	NIFTY—Dhaanya	NIFTY—Comdex	NIFTY—iComdex	NIFTY—Gold (iComdex)	NIFTY—Crude Oil (iComdex)	Number of Observations (max.)
May 2006 to March 2008	Commodity growth cycle associated with double-digit economic growth	0.24	0.39				482
April 2008 to March 2010	Global financial crisis	0.60	0.76				487
April 2010 to July 2013	Struggling recovery	0.11	−0.06	−0.62	−0.38	−0.36	838
August 2013 to September 2015	CTT/NSEL-related confidence crisis	0.57	−0.85	−0.81	−0.71	−0.81	533
May 2006 to December 2012	Period of global co-movement	0.39	0.65	−0.36	0.56	−0.52	1659
April 2010 to September 2015	Stabilised period till merger with SEBI	0.56	−0.28	−0.82	−0.64	−0.71	1371
September 2015 to March 2019	Post-merger with SEBI	0.43	0.87	0.63	0.13	0.25	864
Long period Average May 2006 to May 2019		0.84	0.51	−0.79	−0.48	−0.80	3229

Source: SEBI, MCX and NCDEX.

Note: Data for Dhaanya before October 2010 are back-tested data; the index was renamed the Nkrishi Index in 18 March 2019. For the trial iComdex, data before September 2017 are back-tested data; trial data started from 2 January 2012. The Comdex series was discontinued on 16 January 2020 and replaced with the new iComdex. Back-tested data for the new iComdex series, launched on 20 December 2019, are available only from 31 December 2015.

For the trial iComdex, data before September 2017 are back-tested data; trial data started from 2 January 2012. The Comdex series was discontinued on 16 January 2020 and replaced with the new iComdex. Back-tested data for the new iComdex series, launched on 20 December 2019, are available only from 31 December 2015.

As in the case of global markets, during 2002–2012 (data are available for Indian commodity markets from 2006 onwards), both the agri (Dhaanya) and non-agri (Comdex) segments showed positive correlation with equity markets indicating synchronised movements, despite the total absence of known conditions enabling financialisation in India. Further, it is observed that, after 2015 commodity futures indices tend to be more correlated with the equity index, NIFTY. This is true even for the return-based iComdex series. This co-movement is observed contrary to the global situation, wherein markets have been decoupled after 2012. This confirms that there ought to be some alternate explanations for co-movement, other than financialisation or entry of financial players into commodity markets.

Mirror contracts could be a prime suspect when co-movement is coincident with the global co-movement phase, as Indian equity markets are integrated with global markets through FPI flows. However, post-2015 co-movement requires some other explanation. Commodity markets were brought under the securities market regulator, SEBI, in September 2015; however, the real integration of brokers and exchanges happened only in later years. In any case, nearly the same set of brokers has been operating in both segments, through a wholly owned subsidiary before SEBI’s takeover, and now perhaps through the same entity. Hence, it cannot be the case that the merger with SEBI brought so-called financialisation into the market. Another distinguishing feature of the post-2015 period is the inflation-targeting regime of the RBI, wherein the central bank was given the single-focused mandate to tackle general commodity prices (at the level of consumers) by adjusting interest rates (the repo rate) in the economy.⁷

On the expectation of an increase in commodity prices from the current level, the central bank raises the repo rate, which in turn keeps equity markets down through sentiments as well as the increased cost of leverage. On the expectation of softening commodity prices in future, repo rates are brought down, cheering the equity markets. The price formation of commodity futures is linked to financial markets via the interest rate, as F₀ = S₀ + I + (W - U), where F₀ = futures price at t = 0, S₀ = spot price at t = 0, I = interest cost or the opportunity cost of investing that money, W = warehousing or storage cost and U is the utility of holding inventory/ stock. Thus, it could be through the agency of interest rates that both equity and commodity markets now move together. A simple correlation of the weighted average monthly call rate (which would immediately reflect the intermittent changes in the repo rate or the policy rate) with the monthly average of the NIFTY, and monthly closing prices of Dhaanya and Comdex for the relevant period is given in Table 3.

Table 3.

Correlation of Interest Rates with NIFTY, Dhaanya and Comdex

Correlation Coefficients
Period	Nifty—Call Money	Dhaanya—Call Money	Comdex—Call Money	Number of Observations
April 2010 to September 2015	0.03	0.54	0.78	66
September 2015 to March 2019	−0.50	−0.11	−0.55	42

Source: Dhaanya, Comdex and NIFTY have been sourced from the SEBI (2019, April). The weighted average monthly call rates have been taken from the RBI database.

It is striking to note that all the indices now move in the same direction as the interest rate. Prior to 2015, the NIFTY had no major correlation with the call money rates; after 2015, it has a negative correlation of –0.5. Prior to 2015, both Dhaanya and Comdex had a high positive correlation with interest rate changes, as interest costs get integrated into the prices of commodities. The interest cost is a significant part of the carrying cost⁸ of traders who need to hold stocks for settlement. With the new inflation-targeting regime, which targets expected commodity prices, the relationship has turned negative. In a way, these coefficients also indicate crudely, the impact of inflation-targeting in India. As an exercise in expectation moulding, the influence of inflation-targeting on futures prices rightly captures its effectiveness on tempering expectations. While agri-futures prices seem to be only weakly impacted by interest rate changes now, metals and energy futures prices, as reflected through Comdex, seem to be moderately influenced. Agri prices are ruled mostly by structural, supply-side factors (say crop failure, drought, etc.), and targeting these prices through repo rate changes could be a futile exercise. The above relationship between equity and commodity indices is not found when the correlation of daily returns of the indices is taken (Tables 6 and 7). The commodity markets—both the agri and non-agri segments—still show a weak negative or no correlation with equity markets. An analysis based on phases of business cycles in India also did not reveal any particular association between indices.

It is also observed that specific commodities like gold or crude oil show a near consistent negative relationship with equity markets, save for the positive correlation seen after 2015. An increase in crude oil prices is seen as slowing growth in India affecting its fiscal and external balances⁹ and pulling down the NIFTY, whereas gold retains its safe haven status as an inflation hedge. It appears that after the 2008 global financial crisis, agri-futures prices are consistently positively correlated with equity markets whereas non-agri futures show mostly a negative correlation. Most of the commodities comprising the agri index have industrial usage too (e.g., soyabean, castor seed, barley, guar seed, etc.). If the above relationship between agri derivatives and equity indices holds true in foreign jurisdictions as well, then it is a possibility that when UNCTAD observed co-movement of equity and commodity indices during 2002–2008, it could be mainly because of the agri-price rise around that time, fuelled by global growth, rather than due to the entry of financial players or index investing. As per the data from BIS, only 5% of the OTC contracts were in gold or precious metals in 2008. Thus, instead of generalising that commodity prices would move in an opposite direction to equity, it is better to limit the inverse association to commodities like gold or crude oil.

5. Changing Market Dynamics: Results from Linear Regressions and Rolling Regressions

Since correlation coefficients are not sufficient enough to assert on the impact of inflation targeting or interest rate dynamics, a simple linear regression was fitted on the monthly index data and on other macro-variables of potential influence. However, to avoid non-stationarity problems,¹⁰ variables were transformed to the rate of growth (return) format [(Y_t – Y_t_–1)/Y_t_–1], which are stationary at levels [in other words, integrated of order 0 or I ~ (0)]. In addition to the call rate (Call), the wholesale price inflation (WPI), exchange rate (ER) and deseasonalised Index of Industrial Production (IIPds) were included as macro-variables relevant to determining asset returns (NIFTY, Comdex and Dhaanya). The IIP is considered a good proxy for GDP (which is otherwise not available on a monthly basis), while returns on the call rate and exchange rate represent returns in the money market and foreign exchange market, respectively. Changes in the WPI reflect the month-on-month growth in inflation. (The central bank targets the consumer price index, however, the WPI is the relevant variable for commodity markets.) The exchange rate is represented in USD–INR format, and an increase in the exchange rate reflects a depreciation of the Indian rupee, which could result in increased commodity prices particularly of imported goods like crude oil, gold, etc. The regression is also an attempt to see whether commodity markets reflect general macro-economic fundamentals and returns in other segments of the financial market.

Table 4.
Correlation Matrix for April 2010 to September 2015

Call NIFTY WPI Comdex Dhaanya ER IIPds

Call 1.00

NIFTY 0.03 1.00

WPI 0.66 0.58 1.00

Comdex 0.78 –0.32 0.49 1.00

Dhaanya 0.54 0.56 0.90 0.34 1.00

ER 0.47 0.70 0.91 0.20 0.89 1.00

IIPds 0.42 0.81 0.78 0.09 0.71 0.81 1.00

	Call	NIFTY	WPI	Comdex	Dhaanya	ER	IIPds
Call	1.00
NIFTY	0.03	1.00
WPI	0.66	0.58	1.00
Comdex	0.78	–0.32	0.49	1.00
Dhaanya	0.54	0.56	0.90	0.34	1.00
ER	0.47	0.70	0.91	0.20	0.89	1.00
IIPds	0.42	0.81	0.78	0.09	0.71	0.81	1.00

Source: The author.

Note: Green indicates high positive correlation, red indicates high negative correlation.

Table 5.

Correlation Matrix for October 2015 to March 2019

	Call	NIFTY	WPI	Comdex	Dhaanya	ER	IIPds
Call	1
NIFTY	−0.51	1.00
WPI	−0.39	0.94	1.00
comdex	−0.55	0.88	0.90	1.00
Dhaanya	−0.11	0.45	0.59	0.52	1.00
ER	0.44	0.25	0.49	0.36	0.57	1.00
IIPds	−0.41	0.92	0.94	0.88	0.51	0.41	1.00

Source: The author.

Note: Green indicates high positive correlation, red indicates high negative correlation.

Table 6.

Correlation Matrix (in return form) for April 2010 to September 2015

	rCall	rNIFTY	rWPI	rComdex	rDhaanya	rER	rIIPds
rCall	1.00
rNIFTY	−0.19	1.00
rWPI	0.15	−0.07	1.00
rComdex	0.29	−0.04	0.36	1.00
rDhaanya	0.12	−0.06	0.10	0.26	1.00
rER	0.25	−0.63	0.12	0.10	0.02	1.00
rIIPds	0.12	0.28	−0.03	0.06	−0.11	−0.20	1.00

Source: The author.

Note: Green indicates high positive correlation, red indicates high negative correlation.

The correlation matrix of variables without transformation into the return format, before and after September 2015 are shown in Tables 4 and 5. The correlation matrix shows that all asset prices including the WPI and production index of the IIP have reversed their relationship with the call rate after September 2015. Such a high correlation is not visible when variables are taken in the return form (Tables 6 and 7), though a sign reversal in commodity prices is clearly visible. However, due to time series properties, the variables cannot be used without their transformation in the regression equation. The results of the regression equations, when variables are taken in return form, are given in Table 8, along with other relevant test statistics for checking their robustness.

Table 7.

Correlation Matrix (in return form) for October 2015 to March 2019

	rCall	rNIFTY	rWPI	rComdex	rDhaanya	rER	rIIPds
rCall	1.00
rNIFTY	−0.10	1.00
rWPI	−0.25	0.24	1.00
rComdex	−0.28	0.05	0.34	1.00
rDhaanya	−0.25	0.17	0.48	0.06	1.00
rER	0.28	−0.56	0.19	0.24	−0.13	1.00
rIIPds	−0.22	0.05	0.28	0.24	0.12	0.07	1.00

Source: The author.

Note: Green indicates high positive correlation, red indicates high negative correlation.

The cumulative sums (CUSUM) of the recursive residuals and their squares from the regression specified (for the entire period) showed that the equation of Dhaanya exhibits some stability issues. The best transformation for Dhaanya is in the first difference form (i.e., as Δ, in absolute changes rather than as relative changes). Since the objective here is to put both commodity and equity indices into the same modeling/variable environment, this issue is sidelined for the moment, particularly since there is a return to the confidence band. Further, save for the absolute values of the parameters, the behavioural patterns and results do not change when variables are taken in the first difference form.¹¹ Since financial market players generally respond based on returns, this transformation is adopted for the analysis.

Table 8 shows that even though the call rate is not a significant variable in determining asset returns, the change in its sign over the two periods is unique to the call rate. Prior to 2015, an increase in the interest rate reduced equity market returns and increased commodity price indices or commodity returns, as the interest rate could have been working as a typical line item in the cost of operation. Post 2015, an increase in the interest rate depresses commodity futures price returns or commodity returns, while increasing equity returns on the good news of a potential fall in inflation. The interest rate or exchange rate or general economic mood as reflected in the IIP are not quite sufficient in explaining the price formation in commodity markets, particularly in the pre-September 2015 period. The WPI seems to have some limited influence on commodity returns and post-2015 values of NIFTY. It could be that commodity-specific factors are the main drivers of price formation in commodity derivative markets, given the very low R² values for the equations for Comdex and Dhaanya. The post-2015 coefficients of macro-variables in the equation for the NIFTY attests to the ‘irrational exuberance’ in the equity market during that period. To understand the changing dynamics of these variables in determining asset prices, a rolling regression¹² was done to identify the exact turning point in the relationship between these variables.

Table 8.

Linear Regression Results for the Two Time Periods Before and After September 2015

	NIFTY Returns			Dhaanya Returns			Comdex Returns
	Pre-September 2015	Post-September 2015	Entire Period	Pre-September 2015	Post-September 2015	Entire Period	Pre-September 2015	Post-September 2015	Entire Period
Money market return (changes in call rate)	−0.035	0.33	−0.021	0.109	−0.143	0.084	0.157	−0.588	0.11
Return in forex market (changes in exchange rate)	−1.05**	−1.52**	−1.12**	−0.125	−0.57	−0.167	0.031	0.77	0.123
Month-on-month growth in IIP (deseasonalised)	0.274*	0.034	0.21	−0.307	−0.045	−0.187	0.079	0.206	0.167
Month-on-month changes in WPI	0.107	1.99**	0.56	0.658	2.97**	1.55*	2.12**	1.05	2.01**
Constant	0.013**	0.008**	0.011**	0.015*	−0.0008	0.009*	−0.008	0.002	−0.004
R²	0.43	0.48	0.41	0.037	0.29	0.06	0.19	0.23	0.15
F-test result (overall significance of the equation)	Significant	Significant	Significant	Not significant	significant	Not significant	Not significant	Not Significant	Significant
Durbin Watson (test for autocorrelation: acceptable range 1.5–2.5)	1.74	1.83	1.72	1.52	1.89	1.68	2.47	1.84	2.25
Mean VIF (variance inflation factor for multi-collinearity): acceptable range below 0–10	1.09	1.19	1.07	1.09	1.19	1.07	1.09	1.19	1.07
Breusch–Pagan/Cook–Weisberg test for heteroscedasticity at 1% (to check whether residuals are homoscedastic or have constant variance around the fitted value)	Constant variance	Constant variance	Constant variance ^a	Constant variance^b	Constant variance	Constant variance^b	Constant variance	Constant variance	Constant variance
Omitted variables/specification test (Ramsey test) at 5%	OK	OK	OK	OK	OK	OK	OK	OK	OK^a
No. of observations	65	42	107	65	42	107	65	42	107

Source: The author.

Note: ** and *Represents significance at 5% and 10% level, respectively.

^a Shows heteroscedasticity or specification errors at 10% level of significance.

^b Shows heteroscedasticity at 5% level.

Robust standard errors are used for the estimation.

In the present analysis, a recursive approach was adopted wherein the coefficients were estimated using the above linear regression, first using a window of the first 12 months, which was then stepped up by one month each time, till all 108 months are included{Asset price (i.e., NIFTY/Comdex/Dhaanya) = f(Return in money market [i.e., month on month changes in call rate], return in foreign exchange market [i.e., month-on-month changes in the exchange rate], changes in month-on-month IIP [deseasonalised IIP is used], month-on-month changes in wholesale price index (WPI) calculated for all-commodities)}. The difference in the β values indicates the marginal effect of that additional period on the estimates. The βs so estimated (shown as ‘b_variable name’) are plotted against time to see their movement over time and the structural changes if any. The results were the same with a first window size of 18 months or 25 months (this was done to see whether the shorter time period caused fluctuations in β in the initial phase). The movement of the parameters of the call rate, exchange rate, WPI and IIP in the equations of NIFTY, Comdex and Dhaanya are shown in Figure 6. To better clarify this, on the secondary axis, the movements in the policy repo rate, WPI, IIP and INR–USD exchange rate is also plotted in one panel.

The rolling regression of the NIFTY shows that while the coefficients of the exchange rate and IIP are reasonably stable after the initial periods, a trend reversal occurs in the call rate and WPI. The parameter value changes from positive to negative for the call rates during June 2012, and over April to May 2015, it remains range-bound. This could be due to the increasing use of the interest rate as a principal instrument of monetary policy. It was at the May 2011 monetary policy announcement that the RBI stated its intention to create a policy corridor or interest rate corridor for managing the interest rate. The repo and reverse repo rates [of the Liquidity Adjustment Facility (LAF) window] were being fixed separately till this monetary policy announcement, during which it was also stated that the reverse repo rate would not be announced separately but would be linked to the repo rate by keeping it at a certain basis point (100 basis points = 1%) below the repo rate. Thus, the reverse repo ceased to exist as an independent rate and the repo rate emerged as the policy rate.

The spread between the repo and reverse repo rates forms the lower band of the interest rate corridor or policy corridor. The size of the corridor is not often changed though the rates may vary. However, the width of the corridor has been systematically brought down from 200 to 50 between April 2015 and April 2017 as part of the fine-tuning of strategy (RBI, 2019). This was done with a view to ensure a finer alignment of the weighted average call rate or the overnight money market rates with the repo rate (which essentially means more effective transmission of monetary policy).¹³ This interest rate corridor management has been further streamlined with a formal inflation-targeting policy since 20 February 2015. The kinks in the β values of the call rate in the equation of the NIFTY indeed correspond with these dates of changes in monetary policy. The coefficient of the call rate as well as the WPI presents a stable range bound movement after 2015.

Figure 6

Source: The author.

In the case of commodity indices—the Comdex and Dhaanya—an additional observation is the sudden displacement of the market in July–August 2013, and gradual tightening of βs in a range thereafter. This is the time period during which the payment crisis in the National Spot Exchange Ltd. (NSEL) occurred, consequent to which commodity markets came under greater regulatory scrutiny. Since July 2013, the audit trail has been brought into the market through the imposition of commodity transaction taxes (CTT), which also had the same effect as an interest cost. This looks like a structural displacement of the commodity market towards better behaviour, with indices responding to some of the macro fundamentals in a more systematic manner.

6. Causality: Results from Vector Autoregression Modelling

The above analysis does not factor in the simultaneity or endogenity ingrained in the system. The various segments of the financial market could be moving together with some segments causing some others. In a vector autoregression (VAR) framework, one need not worry about determining which variables are endogenous or exogenous. All variables in the VAR systems are considered as endogenous. Hence, lagged values of all the variables including that of the dependent variable are used in predicting the current value of the dependent variable. The usual OLS methods can be applied to each equation separately, with the only requirement being that in an m-variable VAR model, all the m variables should be jointly stationary. Unlike simultaneous equation models, a VAR model is a-theoretic. The VAR framework helps in determining Granger causality, which is our primary concern here.

In our analysis, each equation contains its own two-lagged values and two-lagged values for all other variables in the system, in addition to a constant and a dummy variable, which is an exogenous variable decided a priori, taking the value of 1 after September 2015 and a value of 0 before that. Variables are taken in the return form [named as r(variable name)] so that they are all stationary at levels. The system consists of seven equations, each containing 16 parameters. Table 9 shows that, on the whole, equations of asset prices are not significant, whereas those of macro-variables—WPI, IIP (deasonsalised), exchange rate and interest rate—are significant and explainable with asset prices. The short-run equations (not shown here) in the VAR model show that 2015 indeed presented a structural break year in respect of the WPI. However, the dummy variable is not significant in any other equation.

Table 9.
VAR Model: Overall Significance of Each Equation

Sample: July 2010 to March 2019

No. of obs. 105

Equation Parameters RMSE R ² F p > F

rNIFTY 16 0.034 0.18 1.280 0.232

rComdex 16 0.045 0.12 0.822 0.651

rDhaanya 16 0.049 0.21 1.549 0.105

rWPI 16 0.005 0.53 6.707 0.000

rIIPds 16 0.018 0.43 4.468 0.000

rER 16 0.017 0.28 2.360 0.007

rCall 16 0.040 0.32 2.769 0.002

Log likelihood 1805.53

FPE 0.00

Det (Sigma_ml) 0.00

AIC −32.26

HQIC −31.11

SBIC −29.43

Sample: July 2010 to March 2019
Equation	Parameters	RMSE	R ²	F	p > F
rNIFTY	16	0.034	0.18	1.280	0.232
rComdex	16	0.045	0.12	0.822	0.651
rDhaanya	16	0.049	0.21	1.549	0.105
rWPI	16	0.005	0.53	6.707	0.000
rIIPds	16	0.018	0.43	4.468	0.000
rER	16	0.017	0.28	2.360	0.007
rCall	16	0.040	0.32	2.769	0.002
Log likelihood	1805.53
FPE	0.00
Det (Sigma_ml)	0.00
AIC	−32.26
HQIC	−31.11
SBIC	−29.43

Source: The author.

Note: The lag length indicated by likelihood ratio is 2, while the AIC SBIC criterion indicated one lag length. However, at one lag length, heteroscedasticity appears; hence, the lag length was taken as 2. *** Represents significance at the 1% level.

Table 10.

Granger Causality Results

Hypothesis		F	p > F
Variables that do not Granger cause NIFTY
	rComdex	3.812	0.026**
	rDhaanya	0.026	0.974
	rWPI	0.385	0.682
	rIIPds	2.206	0.116
	rER	0.708	0.495
	rCall	1.906	0.155
	ALL	1.441	0.163
Variables that do not Granger cause Comdex
	rNIFTY	1.685	0.191
	rDhaanya	0.413	0.663
	rWPI	1.919	0.153
	rIIPds	0.081	0.923
	rER	0.010	0.990
	rCall	0.952	0.390
	ALL	0.952	0.500
Variables that do not Granger cause Dhaanya
	rNIFTY	1.042	0.357
	rComdex	2.543	0.084*
	rWPI	0.825	0.442
	rIIPds	1.431	0.245
	rER	1.025	0.363
	rCall	0.893	0.413
	ALL	1.449	0.159
Variables that do not Granger cause WPI
	rNIFTY	1.195	0.308
	rComdex	12.733	0.000***
	rDhaanya	0.661	0.519
	rIIPds	0.954	0.389
	rER	0.441	0.645
	rCall	1.127	0.329
	ALL	3.152	0.001***
Variables that do not Granger cause IIP (deseasonalised)
	rNIFTY	1.154	0.320
	rComdex	0.663	0.518
	rDhaanya	1.191	0.309
	rWPI	0.011	0.989
	rER	1.292	0.280
	rCall	2.275	0.109
	ALL	1.325	0.219
Variables that do not Granger cause the exchange rate
	rNIFTY	0.467	0.629
	rComdex	3.871	0.024**
	rDhaanya	0.217	0.806
	rWPI	1.862	0.161
	rIIPds	2.566	0.083*
	rCall	0.771	0.466
	ALL	1.877	0.048**
Variables that do not Granger cause the call rate
	rNIFTY	0.921	0.402
	rComdex	3.916	0.023**
	rDhaanya	3.018	0.054**
	rWPI	3.019	0.054**
	rIIPds	0.391	0.677
	rER	1.522	0.224
	ALL	2.293	0.014***

Source: The author.

Note: ***, ** and * Represents significance at the 1%, 5% and 10% level, respectively.

Table 10, presenting the Granger causality, shows (a) individually whether the specific X variable is causing Y and (b) whether all the X variables in the system jointly cause the specified Y variable. The rejection of the null hypothesis confirms the causality.

It is seen that variables Comdex and IIP appear to be the purely independent or exogenous variables in the system, as none of the other variables cause them either individually or together. Further, it is not the NIFTY that makes commodity indices move, but the converse: the Comdex seems to be causing movements in the NIFTY, WPI, exchange rate and call rate (significant at the 5% level) and in Dhaanya (at the 10% significance level). Since there is a 40% weightage in the Comdex for energy products, it could be reasonably presumed that the crude oil price is the single most important price that causes movements in all the macro-variables of concern to us. The interest rate (call rate) seems to be tracking all the asset prices—particularly the Comdex, Dhaanya and WPI. However, the interest rate still does not have a statistically significant influence on the WPI or on commodity futures prices or the NIFTY; as of now, it seems to be successful only in giving a signal to the markets, while asset prices follow their own supply-demand dynamics.

7. Conclusion

This article tries to unravel the drivers of financialisation in global commodity markets and in India. Over the past four decades, coupling of the market and synchronised movements have happened globally mainly in the decade 2002–2012, holding firm through troughs and crests. Thereafter, the markets have decoupled. The mirror contract structure in Indian non-agri commodity markets could have transported this global correlation to Indian markets during this period. The data from India attest to the fact that agricultural futures prices are moderately to highly positively correlated with equity markets, whereas non-agri futures, particularly in gold and crude oil, show mostly a negative correlation. Coupling of equity and commodity markets could potentially be an indication of great uncertainty in the markets. As far as India is concerned, the circumstances that enable financialisation of commodity markets (such as the presence of index investing or hedge fund investors) are either weak or non-existent. However, after 2015, there has been a greater correlation between equity and commodity markets even when decoupling is observed in global markets. The only channel of spillover effects from the equity segment in India comes through the proprietary trading desks of brokers who are active in both the equity and commodity segments and whose participation did not change drastically after 2015 when SEBI took over the regulation of commodity derivative markets. A potential suspect is the inflation-targeting policy, wherein interest rates are changed based on expected commodity prices.

The results from the linear regression show that even though the call rate is not emerging a significant variable in determining asset returns, the change in sign during the pre- and post-2015 periods is unique to the call rate. The results from the rolling regression attest to a shift in the WPI and call rate after 2015. The linear regression framework as well as the VAR framework, which accounts for simultaneity, confirms that markets are moving on their own drivers, and macro-variables like the interest rate, exchange rate, or general economic mood (as reflected in the IIP) are not quite sufficient in explaining the price formation in commodity markets. In nutshell, we cannot find any evidence for financialisation of commodity markets in India in the traditional sense of the term. Further, it is not the NIFTY which makes commodity indices move; rather, it is the other way round. The results from Granger causality show that the Comdex is the single most important factor causing movements in the NIFTY, WPI, exchange rate, call rate and Dhaanya. Since 40% weightage in the Comdex is for energy products, it could be reasonably presumed that the crude oil price is the single most important price that causes movements in all the macro-variables of concern to us. The interest rate (call rate) seems to be tracking all the asset prices—particularly the Comdex, Dhaanya and WPI. However, the interest rate still does not have a statistically significant influence on the WPI or on commodity futures prices (both Dhaanya and NIFTY). As of now, it seems to be successful only in giving a signal to the markets, while asset prices follow their own supply-demand dynamics. The rolling regression also brought to light the disciplining effect of regulatory scrutiny and the audit trail in the Indian commodity markets around July 2013. This may be studied further in a panel setting to circumvent the constraints posed by time series analysis and by extending it to the latest years using new age commodity indices.

Footnotes

Acknowledgements

The author thanks Dr C. P. Chandrasekhar (Retd Professor, JNU) and Dr N. R. Bhanumurthy (Vice Chancellor, Dr. B. R. Ambedkar School of Economics) for their valuable guidance and Dr Shunmugham, MCX; Mr Ravi Varanasi, NSE; and Mr Vijay Kumar, NCDEX for valuable insights and data support. Views expressed herein are personal and do not represent the views of the Government of India.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

Notes

References

Adams

, & Glück

(2015). Financialization in commodity markets: A passing trend or the new normal? Journal of Banking and Finance, 60, 93–111.

Awasthi

, Ahmad

, Rahman

, & Phani

(2020). When US sneezes, clich’es spread: How do the commodity index funds react then. Resources Policy, 69. 10.1016/j.resourpol.2020.101858

Bianchi

R. J.

, Fan

J. H.

, & Todorova

(2020). Financialisation and de-financialisation of commodity futures: A quantile regression approach. International Review of Financial Analysis, 68, 1014–51.

Bicchetti

, & Maystre

(2012). The synchronized and long-lasting structural change on commodity markets: evidence from high frequency data. UNCTAD.

Büyükşahin

, & Robe

M. A.

(2014). Speculators, commodities and cross-market linkages. Journal of International Money and Finance, 42, 38–70.

Gorton

, & Rouwenhorst

K. G.

(2006). Facts and fantasies about commodity futures. Financial Analysts Journal, 62(2), 47–68.

Hamilton

J. D.

, & Wu

J. C.

(2015). Effects of index-fund investing on commodity futures prices. International Economic Review, 56(1), 187–205.

Henderson

B. J.

, Pearson

N. D.

, & Wang

(2015). New evidence on the financialization of commodity markets. The Review of Financial Studies, 28(5), 1285–1311.

Irwin

S. H.

, & Sanders

D. R.

(2011). Index funds, financialization, and commodity futures markets. Applied Economic Perspectives and Policy, 33(1), 1–31.

10.

Kumar

, & Somnath

(2014). Is enough being done to regulate global commodity markets? Economic and Political Weekly, 49(18), 67–71.

11.

Mayer

(2012). The growing financialisation of commodity markets: Divergences between index investors and money managers. Journal of Development Studies, 48(6), 751–767.

12.

Nissanke

(2012). Commodity market linkages in the global financial crisis: Excess volatility and development impacts. Journal of Development Studies, 48(6), 732–750.

13.

Plantier

(2013). Commodity markets and commodity mutual funds. Business Economics, 48(4), 231–245. 10.1057/be.2013.29

14.

RBI. (2019). Report of the internal working group to review the liquidity management framework. Reserve Bank of India.

15.

Rossi

(2012). The changing relationship between commodity prices and equity prices in commodity exporting countries. IMF Economic Review, 60(4), 533–569.

16.

SEBI. (2019, April). SEBI Handbook of Statistics and Monthly Bulletin. SEBI.

17.

Sen

(2014). Dominant finance and stagnant economies. Oxford University Press.

18.

Tang

, & Xiong

(2012). Index investment and the financialization of commodities. Financial Analysts Journal, 68(6), 54–74.

19.

UNCTAD. (2012). Development and globalisation: Facts and figures (Chapter 2). United Nations.

20.

UNCTAD. (2013). Commodities and development report. UNCTAD.

21.

Valiante

, & Egenhofer

(2013). Price formation in commodities markets. Financialisation and beyond (Report of an ECMI/CEPS Task Force). Centre for European Policy Studies.

Sample: July 2010 to March 2019
No. of obs.	105
Equation	Parameters	RMSE	R ²	F	p > F
rNIFTY	16	0.034	0.18	1.280	0.232
rComdex	16	0.045	0.12	0.822	0.651
rDhaanya	16	0.049	0.21	1.549	0.105
rWPI	16	0.005	0.53	6.707	0.000
rIIPds	16	0.018	0.43	4.468	0.000
rER	16	0.017	0.28	2.360	0.007
rCall	16	0.040	0.32	2.769	0.002
Log likelihood	1805.53
FPE	0.00
Det (Sigma_ml)	0.00
AIC	−32.26
HQIC	−31.11
SBIC	−29.43

Financialisation of Commodity Markets: Evidence from India

Abstract

Keywords

1. Introduction

2. Existing Global Evidence

3. Is There a Co-movement of Global Equity and Commodity Indices?

Table 4. Correlation Matrix for April 2010 to September 2015 Call NIFTY WPI Comdex Dhaanya ER IIPds Call 1.00 NIFTY 0.03 1.00 WPI 0.66 0.58 1.00 Comdex 0.78 –0.32 0.49 1.00 Dhaanya 0.54 0.56 0.90 0.34 1.00 ER 0.47 0.70 0.91 0.20 0.89 1.00 IIPds 0.42 0.81 0.78 0.09 0.71 0.81 1.00

Footnotes

Acknowledgements

Declaration of Conflicting Interests

Funding

Notes

References

Table 4.
Correlation Matrix for April 2010 to September 2015

Call NIFTY WPI Comdex Dhaanya ER IIPds

Call 1.00

NIFTY 0.03 1.00

WPI 0.66 0.58 1.00

Comdex 0.78 –0.32 0.49 1.00

Dhaanya 0.54 0.56 0.90 0.34 1.00

ER 0.47 0.70 0.91 0.20 0.89 1.00

IIPds 0.42 0.81 0.78 0.09 0.71 0.81 1.00