Paper to submit final version Research Gate Version

3 downloads 0 Views 137KB Size Report
overall capital inflow boosted from US $ 15 billions to US $ 250 billions in 2009 (Irwin and Sanders, .... FTSE 100 Index ... DAX 30, INDBOMB and FTSE 100.

1

CROSS-CORRELATION ANALYSIS OF FINANCIALIZATION WITHIN INTERNATIONAL MAIN INDUSTRIAL METALS MARKETS To full cite this paper: 2018, A. Focacci, “Cross-correlation analysis of financialization within international main industrial metals markets”. Proceedings of the 7th International Academic Conference on Social Science, Multidisciplinary and Globalization Studies, Madrid 26-29 March, p. 5-16. ISBN: 978-605-82290-7-5

Abstract Current financialization process involving commodity markets spurred controversial issues among policy-makers, practitioners and scholars about spillover effects on the price levels, and inherent consequences on the whole economy. In this debate, it is possible to distinguish between two basic and different positions. On one side are “financialization supporters” advocating the influence of institutional investors’ portfolio management strategies. On the opposite, are all those considering traditional economic factors linked to supply-demand imbalances. In the present paper, with the aim to contribute to the discussion, a cross-correlation function (ccf) is applied between Stock Exchange Indexes and main international quoted industrial prices to investigate the lead-lag relationship resulting from the new potential asset linkages. In order to propose a wide analysis, data pertaining some industrialized Countries (Germany, United Kingdom and United States) as well as some important developing Countries (Brazil, China and India) are processed. JEL codes: C32 Time series models D84 Expectations/Speculations G12 Asset Pricing

Keywords: financialization, metal markets, cross-correlation function

2

1. Introduction Commodity markets experienced remarkable changes during the past twenty-years. The most recognized -generally labeled by the term “financialization”- has been characterized by an increasing marketization, financial deregulation and global economic liberalism fueling the augmented participation of newcomer institutional investors like for example Hedge Funds (HF), Commodity Index Funds (CIFs), Commodity Index Traders (CITs; long-only investors like pension funds and insurance companies) and Exchange Traded Funds (ETFs) (Büyükşahin and Robe, 2014). It is estimated the number of CITs more than quadrupled from 2000 to 2010 (Cheng et al., 2015). The number of US HFs more than tripled between 2004 and 2007 (Domansky and Heath, 2007), and overall capital inflow boosted from US $ 15 billions to US $ 250 billions in 2009 (Irwin and Sanders, 2011). Baffes and Haniotis (2016) update the overall amount invested in commodities to December 2015 well over US $ 300 billions. At the same time, the boom of energy and raw material world demand –as deriving from the economic growth of Emerging / Developing Countries (mainly Brazil, China and India) (Focacci, 2005 and 2007)- exerted an unexpected huge pressure on prices (despite the relevant roles played by these Countries also within the supply-side as producers). Empirical researches on financialization effects on commodity markets do not share the same conclusions. One strand of literature highlights the amplified induced volatility as deriving from inherent herding behavior by speculators (among others Gabaix et al., 2006; Engle and Rangle, 2008; Gilbert, 2010; Henderson et al., 2015), and (generally) fewer restrictions new operators have to respect compared with traditional specialists (Rahi and Zigrand, 2009; Teo, 2009). On the opposite side, emphasizing traditional stabilizing action of the speculation mechanisms (Friedman, 1953), are -for example among others- empirical works by Irwin et al. (2009), Stoll and Whaley (2010) and Miffre and Brooks (2013). Following such premises, the main purpose of the present study is to investigate prices dynamics by exploring potential lead-lag effects transmitted from the traditional financial markets (in our case proxied by Stock Exchanges indexes) to main metal quotations as a consequence of specific financial portfolio strategies aiming at managing futures on industrial metals as a conventional financial asset class.

2. Methodology and data To analyze such a research issue, firstly we have to introduce the mechanism hypothesized to igniting the process. In detail, as well depicted by Adams and Glück (2015), “to compensate for a decline in stock prices, investors may reduce their commodities position and invest the proceeds in stocks. A fall in stock prices therefore transmits to the commodity market by reducing commodity prices. Similarly, an increase in stock prices induces investors to sell part of their stock holdings to back their commodity position”. Thus, resulting from the more sophisticated portfolio management strategies implemented by the multitude of operators within (highly) integrated markets, an induced dynamic relationship starting from the (lower or higher) level of stocks’ prices would be finally transmitted to metal quotations (respectively pushing down or up prices) through a merely financial mechanism based on commodities futures indexes. With the aim to investigating this hypothesis, the lagged ccf of monthly quotations of stock indexes and metal prices is processed considering its effectiveness in the detection of the ordinal sequence of events (Warner, 1998). As a matter of fact, by this time-series analysis technique, we do not ascertain a causal relationship between variables (this kind of exercise is usually approached in econometrics by implementing a specific Granger causality test), but we try to trace out their statistical evidence. The procedure is composed by two steps: an initial pre-whitening phase , and a subsequent application of the ccf (calculated between the pair of the residuals) to detect meaningful statistical values (those exceeding the conventional two-standard error 95% upper and lower confidence statistical limits). The pre-whitening phase to fitting original data is carried out by adopting the general ARIMA model usually formalized in mathematical symbolization for nonstationary processes as: Φ (B) (1-B)d Yt = σ + Θ (B) ετ

3

with: - Φ (B) as the Autoregressive (AR) operator; - B as the backshift operator defined as BdYt = Yt-d ; - d as the order of differentiation; - σ as the constant term relating to the mean of the stochastic process; - Θ (B) as the Moving Average (MA) operator. Henceforth, obtained the residuals of the series x and y, data are then processed by computing the ccf at lag k, according to Box et al. (2016), defined as: ccf (k) =

where

with



for k = 0, ± 1, ± 2, …..

is the cross-covariance coefficient at lag k obtained as:



representing



(

− ̅ ) (

− ) for k = 0, + 1, + 2, …..



(

− ) (

− ̅ ) for k = 0, - 1, - 2, ......

standard deviations of the series x and y respectively.

It must be pointed out that ccf is not characterized by a symmetric behavior about 0, and its inherent properties can be effectively exploited to detect whether the response variable y is “reacting” to the explanatory variable x after a (statistical significant) lag time (ccf values for k > 0). An anticipatory behavior could be explored for ccf values in k < 0 side. Furthermore, to investigate anticipatory effects, it is always possible to invert x with y and considering ccf values for k > 0 (Warner, 1998). In the present work, stock exchange log-indexes are assumed as the explanatory variable (x) while logindexes of metal prices as the response variable (y). Following this framework, only the k > 0 side of the ccf positive values has to be considered to investigate a (potential) financialization influence transmitted from x to y. As far as analyzed dataset is concerned, we consider three among more industrialized and financially advanced Countries (Germany, United Kingdom and United States) representing core-established economies. At the same time, three main developing Countries (Brazil, China and India) are selected for new growing ones. More in detail, the overall list of monthly stock indexes quotations for the six Countries (transformed into logarithm form), coming from Datastream (2017), includes: -Brazil Ibovespa stock index (IBOVES) from January 2000 to September 2017; -China Shangai A stock index (CHISHA) from January 2000 to September 2017; -Germany DAX 30 stock index (DAX30) from January 2000 to September 2017; -India S&P Bombay stock index BSE 100 (INDBOMB) from January 2000 to September 2017; -United Kingdom FTSE 100 (FTSE 100) from January 2000 to September 2017; -USA New York stock exchange index (NYSE) from January 2000 to September 2017. For what concerns main industrial metal price series, also in this case gathered from Datastream (2017), they are not taken in the original values, but after a pre-transformation into index form to homogenize elaboration with stock indexes. For this task, the following formula is adopted: Indext = It-1 × (1 +





)

assuming I0 = 100 as the starting value of the series to center the calculations.

4

As well as for Stock indexes, a log-transformation of the indexed-metal quotations is subsequently applied. Industrial metals collected in this study are: Aluminum (Al), Copper (Cu), Lead (Pb), Nickel (Ni) and Zinc (Zn) in their London Metal Exchange (LME) quotations for the same time span already adopted for Stocks. Finally, all time-series are taken beginning on January 2000, assuming such a month as the (conventional) widespread financialization starting point, even if an undisputable and specific date is not fixed in literature (Cheng et al., 2015; Cheng and Xiong, 2014; Irwin and Sanders, 2011; Domansky and Heath, 2007). Descriptive statistics concerning all the log-transformed series are reported within following Table 1.

Table 1 Descriptive statistics for Stock Exchanges and industrial metals log-indexes BRA CHI GER IND Ibovespa Index

Shangai A

Dax 30 Index

Bombay S&P Index

Mean

10.51

7.75

8.77

8.15

Median

10.79

7.74

8.77

8.44

Minimum

9.12

6.97

7.79

6.63

Maximum

11.18

8.71

9.46

9.25

Std dev

0.61

0.37

0.37

0.77

Skewness

-0.79

0.12

-0.23

-0.52

Kurtosis

-0.83 UK

-0.47 USA

-0.44 LME Al

-1.08 LME Cu

FTSE 100 Index

NYSE All Index

Index

Index

Mean

8.63

8.97

4.68

5.50

Median

8.67

8.97

4.66

5.75

Minimum

8.18

8.44

4.29

4.30

Maximum

8.93

9.40

5.18

6.29

Std dev

0.17

0.24

0.22

0.61

Skewness

-0.66

-0.13

0.34

-0.68 -1.01

Kurtosis

-0.39

-0.87

-0.79

LME Pb

LME Ni

LME Zn

Index

Index

Index

Mean

5.62

5.12

5.01

Median

5.91

5.16

5.13

Minimum

4.46

4.00

4.19

Maximum

6.69

6.47

5.98

Std dev

0.64

0.49

0.45

Skewness

-0.64

0.17

-0.23

Kurtosis -1.03 -0.33 Source: Personal elaborations on Datastream (2017); N = number of obs = 213; Monthly frequency (2000:01-2017:09).

-0.89

3. Elaborations and findings In the first step of pre-whitening procedure ARIMA best fitting models are estimated, and all calculated parameters are the following: BRA IBOVES (0,1,0), CHI SHA A (0,1,13), DAX 30 (0,1,0),

5

IND BOMB (0,1,0), FTSE 100 (0,1,0), NYSE (1,1,0), LME Al (0,1,3), LME Cu (1,1,0), LME Pb (0,1,9), LME Ni (0,1,0), and LME Zn (0,1,0). Pre-whitened residuals correlograms (before ccf elaborations) reporting AutoCorrelationFunctions (ACF) are not included here for brevity reasons. Author is available, if needed, to furnish graphs (k = 24) to show residuals after this pre-whitening phase. However, to briefly depict such outcomes, it is possible to evidence no ACF values exceeding statistical upper-lower bounds for Al, Ni, CHISHA, DAX 30, INDBOMB and FTSE 100. One very small exceeding spike is indeed evidenced both in Cu and Pb series. One exceeding (even if not with excessive value) spike for IBOVES and NYSE models. One out of two spikes left in Zn series having a more significant statistical value. Overall, we can reasonably assume to have obtained independent and random distributions. Normality is not tested here because, this is not an inferential exercise. The second step involving ccf elaborations is processed by a time lag k selected equal to ±7 months. Subsequent Fig. 1 exhibits all the ccf for the various combinations. On the x-axis lag k is represented both for negative and positive values specifying, respectively, an anticipatory and a reacting behaviour. The k = 0 value is located in the median position just between negative (on the left) anticipatory and positive (on the right) lagged values. Conventional two-standard error 95% upper and lower confidence statistical limits are reported with solid lines.

IBOVESPA-Cu

.250

.250

.050

.050

ccf

ccf

IBOVESPA-Al

-.150

-.150

-.350

-.350

IBOVESPA-Pb

.250

.250

.050

.050

ccf

ccf

IBOVESPA-Ni

-.150

-.150

-.350

-.350

Fig. 1 Cross-correlation function with 7 lags for the various Stock-Metal Indexes (continued on the next page) Source: Personal elaboration on Datastream (2017)

6

CHI SHA A-Al

.250

.250

.050

.050

ccf

ccf

IBOVESPA-Zn

-.150

-.150

-.350

-.350

CHI SHA A-Ni

.250

.250

.050

.050

ccf

ccf

CHI SHA A-Cu

-.150

-.150

-.350

-.350

CHI SHA A-Zn

.250

.250

.050

.050

ccf

ccf

CHI SHA A-Pb

-.150

-.150

-.350

-.350

DAX 30-Cu

.250

.250

.050

.050

ccf

ccf

DAX 30-Al

-.150

-.150

-.350

-.350

Fig. 1 Cross-correlation function with 7 lags for the various Stock-Metal Indexes (continued on the next page) Source: Personal elaboration on Datastream (2017)

7

DAX 30-Pb

.250

.250

.050

.050

ccf

ccf

DAX 30-Ni

-.150

-.150

-.350

-.350

INDBOMB-Al

.250

.250

.050

.050

ccf

ccf

DAX 30-Zn

-.150

-.150

-.350

-.350

INDBOMB -Ni

.250

.250

.050

.050

ccf

ccf

INDBOMB-Cu

-.150

-.150

-.350

-.350

INDBOMB-Zn

.250

.250

.050

.050

ccf

ccf

INDBOMB -Pb

-.150

-.150

-.350

-.350

Fig. 1 Cross-correlation function with 7 lags for the various Stock-Metal Indexes (continued on the next page) Source: Personal elaboration on Datastream (2017)

8

FTSE 100-Cu

.250

.250

.050

.050

ccf

ccf

FTSE 100-Al

-.150

-.150

-.350

-.350

FTSE 100-Pb

.250

.250

.050

.050

ccf

ccf

FTSE 100-Ni

-.150

-.150

-.350

-.350

NYSE-Al

.250

.250

.050

.050

ccf

ccf

FTSE 100-Zn

-.150

-.150

-.350

-.350

NYSE-Ni

.250

.250

.050

.050

ccf

ccf

NYSE-Cu

-.150

-.150

-.350

-.350

Fig. 1 Cross-correlation function with 7 lags for the various Stock-Metal Indexes (continued on the next page) Source: Personal elaboration on Datastream (2017)

9

NYSE-Zn

.250

.250

.050

.050

ccf

ccf

NYSE-Pb

-.150

-.150

-.350

-.350

Fig. 1 Cross-correlation function with 7 lags for the various Stock-Metal Indexes Source: Personal elaboration on Datastream (2017)

4. Discussion and Conclusions From the above results, and following the interpretation of ccf graphs, we resume outcomes in the following Table 2.

Table 2: Lag value of significant statistical spikes in the ccf analysis of the Stock-Metal Indexes relationship

BRA CHI GER IND UK USA

Al -5, -2,|-1|, 0 No Sig. |0|, 3 -5, -2 , |-1| |0|, 3 |0, 3

Cu |-1|, 0, 3 |-1|, 1,3 |0|, 3 -4, -2, |-1| |0|,3 0

Ni - 1, |0|, 3 |-1| 0 |-1|, 0, 3 |0|, 1 -3,|0|, 1

Pb -3, |0| 0, |5| 1 -3, -2, |0| -3, |1| 1

Zn | -1|,0 |-1|, 5 |0| ,3 |-1|, 2 -5, -4, |0|, 3, 6 |0|, 3

Source: Personal elaboration on Datastream (2017). Values in modulus | | represent the most significant spike

To read the Table, and taking the Ibovespa (Brazil) coupled with Al as an example, it is possible to show -looking at the very first graph presented in Fig. 1- that 4 statistical significant peaks (exceeding the 95% upper and lower confidence levels) in ccf values are present (lags: -5, -2, -1 and 0). The most significant among them is at lag -1 (the highest histogram bar in the diagram, and within Table 2 marked within | | ). The same way to read data can be applied to all further cases. As can be seen from the overall findings, it is not possible to evidence a clear lead-lag effect (from Stocks-to-Metals) in quotations originated from a marked financial induced influence. If this were the case the k > 0 values should be, generally, the statistically meaningful values (represented by the highest bars). In our findings, we have many 0 values (representing no lead-lag effect at all) as well as negative k values (expression of the opposite influence). However, similarities are present considering the three industrialized Countries (see for example Al, Cu, Pb and Zn). This fact could point to common financial path among investors in these Countries, as previously mentioned in the Introduction, but without any evidence that systematic financial portfolio strategies (acting through selling stocks and buying metals) could exert a real influence on quotations. Admittedly, only for Pb

10

and in the cases of Germany and USA (k = 1, as the most significant peak) it could be detected a clear signal transmitted in one month-lag from stocks to lead quotations. Some drawbacks and further implementation steps could be considered to refine and develop this research. For example, one issue could regard the frequency of data. On this point, some critical observers could argue that monthly quotations are not the right periodicities to investigate due to the inherent speed of the phenomenon. This is a good point to develop, nonetheless in this industrial metals first study, the main goal was to investigate, whether a “clear” financial (and artificially) induced mechanism could be detected. Reasonably, the more the phenomenon is affecting the economy as a relevant factor, the more its related effects should be observable. Our present choice in adopting monthly quotations (instead of daily ones, for example) is to avoid the increasing likelihood of finding causal relationships with higher frequency data (Schwarz and Szakmary, 1994). However, considering that financialization effects should be well persistent within markets (due to the hypothesized systematic action of the increased presence of newcomer institutional investors as several observers advocate), it would be hard to foster its consequences could be observable merely under well-defined (and spurious?) conditions. Furthermore, skepticals could not be persuaded for the number and the choice of the Countries analyzed and proposed. Nevertheless this point is solved by arguing that Germany, UK and USA are the more advanced Countries within international financial system (related financial industries are without doubt the more sophisticated around the World). For what concerns the Developing economies, it is important to point out that Brazil, China and India exert an ever growing role in the worldwide industrial metal markets -at least- for their demographic impact. Thus, if something is going on, it seems reasonable to monitor a (whatsoever) trace also within such markets. Another potential drawback could lie in the fact that not all commodities could be affected by financialization mechanisms at the same intensity and time. This aspect could be “easily” addressed in future research works by investigating further commodities. A previous work by Focacci (2017) elaborated and adopting the same ccf methodology (involving international oil markets) supports the same conclusions as derived by the present work for industrial metals. Without any pretension to be exhaustive or definitive, our results do not detect statistical evidences of a meaningful impact of financialization investors’ strategies on metals prices. As previously pointed out, if these assets (and/or their derivatives) are included in common (and systematical) financial management styles, relevant and coherent movements of quotations should be detectable within such a related-to-Stocks scheme. If this were not true, financialization era would appear as a merely different (and probably more sophisticated) market framework contributing -but not affecting in a decisive manner- to price dynamics. Are financialization effects really detectable within international industrial metals markets ?

11

REFERENCES Adams, Z. and Glück, T., 2015. Financialization in commodity markets: a passing trend or the new normal ? Journal of Banking& Finance 60, 93-111.

Baffes, J. and Haniotis, T., 2016. What explains agricultural price movements ? Journal of Agricultural Economics 67 (3), 706-721. Box, G. P. Jenkins, G. M. Reinsel, G.C. and Ljung, G. M., 2016. Time Series Analysis-Forecasting and Control 5th ed. John Wiley & Sons, Hoboken. Büyükşahin, B. and Robe, M.A., 2014. Speculators, commodities and cross-markets linkages. Journal of International Money and Finance. 42, 38-70. Cheng, I.H. and Xiong, W., 2014. Financialization of Commodity Markets. Annual Review of Financial Economics 6, 419-441. Cheng, I.H., Kirilenko, A. and Xiong, W., 2015. Convective risk flows in commodity futures markets. Review of Finance 19 (5), 1733-1781. Datastream, 2017. Accessed in October 2017. Domansky, D. and Heath, A., 2007. Financial investors and commodity markets. BIS Quarterly Review, March 53-67. Engle, R.F. and Rangel, J., 2008. The spline-GARCH model for low-frequency volatility and its global macroeconomic causes. Review of Financial Studies 21 (3), 1187-1222. Focacci, A., 2005. Empirical analysis of the environmental and energy policies in some developing countries. The cases of Brazil, China and India. Energy Policy 33 (4), 543-554. Focacci, A., 2007. Empirical analysis of the relationship between total consumption-GDP ratio and per capita income for different metals. The cases of: Brazil, China and India. International Journal of Social Economics 34 (9), 612-636. Focacci, A., 2017. Is financialization effect detectable within international oil markets? Proceedings of the 5th International Academic Conference on Social Science, Multidisciplinary and Globalization Studies, Vienna 12-14 September, p. 15-26. Friedman, M., 1953. The case for flexible exchange rates. In: Essays in positive economics, Friedman, M. (ed). University of Chicago Press, Chicago. Gabaix, X., Gopikerishman, P., Plerou, V. and Stanley, H.E., 2006. Institutional investors and stock market volatility. Quarterly Journal of Economics 121 (2), 461-504. Gilbert, C.L., 2010. How to understand high food prices. Journal of Agricultural Economics 61 (2), 398-425. Henderson, B.J., Pearson, N.D. and Wang, L., 2015. New evidence on the financialization of commodity markets. Review of Financial Studies 28 (5), 1285-1311. Irwin, S.H. and Sanders, D.R., 2011. Index funds, financialization and commodity futures markets. Applied Economic Perspectives and Policy. 33 (1), 1-31. Irwin, S.H., Sanders, D.R., and Merrin, R.P., 2009. Devil or angel ? The role of speculation in the recent commodity price boom (and bust). Journal of Agricultural and Applied Economics 41 (2), 377-391. Miffre, J. and Brooks, C., 2013. Do long-short speculators destabilize commodity futures markets? International Review of Financial Analysis 30, 230-240. Rahi, R. and Zigrand, J.P., 2009. Strategic financial innovation in segmented markets. Review of Financial Studies 22 (8), 2941-2971. Schwartz, T.V. and Szakmary, A.C., 1994. Price discovery in petroleum markets: arbitrage, coitegration, and the time interval of analysis. Journal of Futures Markets 14 (2), 147-167. Stoll, H. and Whaley, R., 2010. Commodity index investing and commodity futures prices. Journal of Applied Finance 20 (1), 7-46. Teo, M., 2009. The geography of hedge funds. Review of Financial Studies 22 (9), 3531-3561. Warner, R. M. 1998. The Spectral Analysis of Time-Series Data. The Guilford Press, New York.