Financial Contagion and Oil Crisis : A Factor Analysis

Introduction

Several crises such as the subprime crisis, the global financial cri- sis, the 1987 US crash turned upside down the international stock markets. More recently, an oil crisis characterized by a price draw- down, as a consequence of both declining demand and stagnant production, affected negatively the stock markets remerging the financial contagion spectrum. What about financial contagion, is it financial linkage across stock markets or shock transmission from one market to another as a virus transmission? How financial contagion could be detected? Which indicators could be used to distinguish between interdependence and shock transmission? The scope of this paper is to test financial contagion resulting of the actual oil crisis applying a component analysis, for this purpose, moult indicators were computed, such as KMO, proportion of variance explained by the first component and number of pertinent factors for periods before and during the oil crisis, considered as quiet and turbulent periods. The paper is organized as follows, a first section will define the financial contagion concept and present a selected literature review about contagion tests based upon com- ponent analysis, then a second section will present the data and methodology, a third section is devoted to main results and a last section will review main findings and point out further researches.

Literature Review

According to Rigobon, contagion between stock markets is defined as shock transmission being stronger than expected on the basis of economic fundamentals (pure contagion), or as shock transmission having suddenly strengthened at times of financial crises (shift contagion) [1]. According to Forbes and Rigobon, contagion is a significant increase of cross market linkages after a shock [2]. Several approaches were implemented to evaluate the financial contagion such as random matrix theory (L. Sandoval Jr and others 2010), wavelett analysis, econometric techniques based upon ARCH and GARCH or copula approach [3-8]. The literature review will focus only on research based upon PCA (principal component analysis) technique testing for financial contagion subsequent to financial crisis.

According to L. Sandoval Jr a crisis is defined to be a crash of global stock markets [9]. The authors count the number of occur- rence of major drops in order to identify major and minor crisis. They identified the following crisis; The crisis of 1987, the 1989 USA saving and loan crisis, 1997 Asian financial crisis, 1998 Rus- sian crisis, 2000 and 2001 end of the dot-com bubble, and 2008 subprime mortgage crisis in the USA. Fuentes and Godoy exam- ined the sovereign spread co-movements for 18 emerging market economies located in Asia, East Europe and Latin America from September 1997 to November 2002 (divided into eight periods) in the light of PCA analysis [10]. They proposed this taxonomy for the co-movements episodes, an extreme Coupling (when the Percentage of variance explained by first principal component is above 50%),a strong Coupling (when the Percentage of variance explained by first principal component is between 35-50%) and a weak coupling otherwise. Following this classification, five epi- sodes of strong coupling were identified the Asian crisis, the Bra- zilian devaluation, the Russian default, the Turkish crisis and the Argentina default.

Silvestre and villar aims to detect financial contagion during the 2007 financial subprime crisis [11]. The data were composed of daily stock prices of 21 developed (OECD) countries during the 2004-2011 period. The period was divided into periods, a quiet one from 1 January 2004 to 31 july 2007 and a turbulent one from 1 august 2007 to 4 March 2011.The PCA analysis revealed differ- ences between both periods in terms of number of relevant factors , three for the quiet one and two for the turbulent one and in terms of proportion of variance explained by the relevant factors, for the quiet one the proportion represents 69,8% while for the turbulent one it reaches 73,9%. M. Ramírez and C. Martínez tries to evaluate the event shock transmission among eight Latin American stock markets [12]. The daily data gathered through 1997 to September 2008 concerned stock market indexes for eight Latin American countries, exchange rates, short term interest rate and sovereign spreads. The authors identifies eight episodes, 4 positive shocks (NASDAQ crisis, Turkish devaluation, Argentina's debt default and Brazilian confidence crisis) and four negatives ones (Mexi- co's debt upgrade, Colombia's debt upgrade, Peru's debt upgrade and Brazil's debt upgrade). then they implement PCA analysis for the subsamples composed of periods before and after shocks. The level of the KMO indicator for the aftershock period appears to be higher than the cut-off point of 0,7 showing a great interdepen- dence. The proportion of variance explained by the first compo- nent considered as an indicator of regional risk has been computed for the aftershock period for each episode. For positive shocks, the proportion of variance explained by the first component is higher than 50% (more than 70% for three episodes) while for negative shocks, the proportion ranges from 40% to 49%.Based upon the threshold proposed by Fuentes and Godoy, the authors conclude to the existence of financial contagion for positive shocks and not for negative shocks [10].

S Mollah and G Zafirov and AMM S Quoreshi wanted to test the presence of financial contagion during the global crisis [13]. The period of analysis was from September 1, 2008 to December 31, 2009. Daily data related to sixty-four MSCI indices were collect- ed grouping nine regions. For each region, the proportion of vari- ance explained by the first component (through PCA analysis) was calculated. The results showed that for the Pacific Asian , South America, North America, Western Europe, Eastern Europe And the Nordic regions, the proportion is greater than 50%, (the highest value of 82% is obtained for North America) while for the MENA, Africa and South and Central Asia the proportion falls below 50% (the lowest value of 32% for the MENA region).The study demon- strates disparity in shock transmission among developed, emer- gent and underdeveloped regions G. Merick , S. Lentz, W. Smeltz and I. Meric aims to study U.S., Latin American, European, and Australasian stock markets linkage during the post-crash period from October 8, 2007 to July 26, 2010 on a weekly basis [14]. Using a PCA technique, they found that the proportion of variance explained by the first component was 66.596%.Two components are able to replicate the whole world (with 74.344% proportion of variance), with a first component composed of U.S., Europe- an, ,Latin American, Australian, Japanese, and New Zealand stock markets while the second component groups Asian stock markets.

Data and Methodology

The data consists in daily closing prices of the following stock market indexes :FTSE 100, Standard and Poor’s500, Nikkei, hang Seng, DAX, CAC40, S&P/CITIC300, BSE Sensex 30, Bovespa, FTSE SOUTH AFRICA, FTSE SINGAPORE, IPC MEXICO, BURSATIL, S&P/ASX 200, Tadawul All Share, Bahrain all share and Brent crude oil. They were extracted from the website fr.in- vesting.com. Returns were computed using the relative difference between two consecutive days. The missing values are replaced by average values during the period. The study period ranges from 3/01/2012 to 25/05/2016.It has been divided into two sub peri- ods :from 3/01/2012 to 01/08/2014 titled before the oil crisis and from 04/08/2014 to 25/05/2016 titled during the oil crisis. The PCA technique will be applied for both sub periods using SPSS software to display the results. The study attempt to analyze the correlations between Brent crude oil price and stock markets pric- es and to compare the pre and post oil crisis periods in terms of co-movements intensity. (coupling) The research hypotheses are the following:

H0: the oil crisis is considered as an extreme episode generating financial contagion

H1: KMO during the oil crisis is greater than the one before the oil crisis

H2: the number of component factors is lower during the oil crisis than before

H3: the proportion of variance explained by the first component factor is greater during the oil crisis than before

Results

Correlation Between Brent Crude Oil Price and Stock Market Returns

Figure 1: Brent Crude Oil Price During the Whole Period

The brent crude oil price evolution shows first a period of stability that lasts around September 2014 with a price of 91,16 $ and the drop will start since October 2014 , to reach in may 2016, 48,45 $.

                                                       Table 1: Correlation Between Stock Market Index Return and Brent Crude Oil Return

Before the oil crisis,7 stock market indexes (FTSE100, SP500, DAX, CAC40, Bovespa, FTSE South Africa and FTSE Singapore) show a significant correlation with brent oil while during the oil crisis, this number increases to 9, adding BSE and IPC Mexico.

PCA Results Before the Oil Crisis (Quiet Period)

Only FTSE100,DAX, CAC40, FTSE SINAGPORE and FTSE South Africa seem to be correlated. Few correlation coefficients are higher than 0,5 .The factorization seems at first sight inadequate. (see appendix 1 for correlation matrix)

The PCA will now be applied to the following variables FTSE100 ,SP500, Nikkei , hang seng, DAX CAC40, SP/CITIC300, Bovespa, FTSE SOUTHAFRICA and FTSE SINGAPORE for the purpose to get higher KMO and significant factors.

                                                                        Table 2: KMO Index and Bartlett Test Quiet Period

The KMO increases but it remains a median value while the Bartlett test authorized the factorization

                                                                                         Table 3: Eigen values Quiet Period

The number of components is of four according to the eigenvalue criteria (cut off point 1) and the proportion of variance explained with four components equals 63,24% while the proportion of variance explained for the first component equals 26,7%

                                                              Table 4: Correlation Between Variables and Components Quiet Period

• Factor 1 is composed of FTSE SOUTH AFRICA, FTSE100,DAX and CAC40 (Europe and South Africa)

• Factor 2 is composed of FTSE SINGAPORE, Bovespa and S&P/CITIC300 (China, Singapore and Brazil)

• Factor 3 is composed of Nikkei (Japan) • Factor 4 is composed of S and P 500 (USA)

PCA Results During the Oil Crisis (Turbulent Period)

The number of correlation coefficients higher than 0,5 is more important than before the oil crisis( see appendix 2 for correlation matrix)

                                                                                    Table 5: Selected Correlation Coefficients

For a selected number of cross-country correlation, the correlation during the crisis appears to be higher than before the crisis.

                                                                                  Table 6: KMO and Bartlett's test Turbulent Period

The KMO is close to 0,76, which depicts a good factorization level and it is validated by the Bartlett test (with a p value equals to 0).

                                   Table 7: Proportion of Variance Explained by Components and Number of Components Turbulent Period

The number of factors retained according to the cut off eigen value point (of 1) is three. The total variance explained by the first component equals 35,042% and 61,288% for the three components.

                                                        Table 8: Correlation Between Variables and Components Turbulent Period

• Factor 1 is composed of FTSE100 DAX CAC40 and FTSE SOUTH AFRICA and SP500 (Europe, USA and South Africa)

• Factor 2 is composed of hang seng and S&P/CITIC300 (China)

• Factor 3 is composed of Nikkei (Japan)

                                                       Table 9: Interdependence Indicators Before and During the Oil Crisis

H1, H2 and H3 are validated due to higher KMO and proportion of variance explained by the first component and to lower number of relevant components during the oil crisis.

According to Fuentes and Godoy (2005) the proportion of variance explained by the first component during the oil crisis is between 35% and 50%, indicating a strong coupling rather than an extreme coupling, so the H4, presence of financial contagion during the oil crisis, is rejected

                                                                                            Table 10: Economic Signifiance of Factor

The oil crisis modifies the component structure, during the oil crisis, USA integrates the first component while before the oil crisis, it monopolizes a whole factor. During the oil crisis, the component structure is more concentrated.

Conclusion

The main findings of this research study are that both KMO, pro- portion of variance explained by the first component and number of pertinent factors play the role of financial contagion indica- tors. The proportion of variance explained by the first component during the oil crisis denotes that the oil crisis is just a strong coupling episode. The KMO during the oil crisis is close to 0,8 which points out good factorization level. The number of components is lower during the oil crisis than before offering also comprehensive factors. The conclusion about contagion presence after the oil cri- sis is not so obvious, whereas the stock market interdependence growth after the oil crisis is undeniable.

Regarding the rule highlighted by Fuentes and Godoy, three re- marks have to be raised, first, The episode tautology (extreme, strong or weak coupling) must integrate other criteria such as the crisis intensity (minor or major), the region stock market index- es (developed, emergent and developing regions) and the type of assets (stock, bonds, commodities,…) second, others principal component indicators have to be added, third , the threshold has to be expressed in terms of difference between the quiet and the turbulent period.

Further researches are expected to test financial contagion during crisis from an investor perspective, considering that an investor must enter the stock market during quiet period and exit from it during turbulent period, joining in that the study conducted by M.Kritzman [15-26].

References

1. Rigobon, R. (2002). Contagion: how to measure it?. In Preventing currency crises in emerging markets (pp. 269- 334). University of Chicago Press.
2. Forbes, K. J., & Rigobon, R. (2002). No contagion, only interdependence: measuring stock market comovements. The journal of Finance, 57(5), 2223-2261.
3. Loh, L. (2013). Co-movement of Asia-Pacific with European and US stock market returns: A cross-time-frequency analysis. Research in International Business and Finance, 29, 1-13.
4. Ranta, M. (2013). Contagion among major world markets: a wavelet approach. International Journal of ManagerialFinance, 9(2), 133-149.
5. Chou, M. R. Y. T., Ng, M. V., & Pi, L. K. (1994). Cointegration of international stock market indices. International Monetary Fund.
6. Hamao, Y., Masulis, R. W., & Ng, V. (1990). Correlations in price changes and volatility across international stock markets. The review of financial studies, 3(2), 281-307.
7. Billio, M., & Caporin, M. (2010). Market linkages, variance spillovers, and correlation stability: Empirical evidence of financial contagion. Computational statistics & data analysis, 54(11), 2443-2458.
8. Rodriguez, J. C. (2007). Measuring financial contagion: A copula approach. Journal of empirical finance, 14(3), 401- 423.
9. Junior, L. S., & Franca, I. D. P. (2012). Correlation of financial markets in times of crisis. Physica A: Statistical Mechanics and its Applications, 391(1-2), 187-208.
10. Fuentes, M., & Godoy, S. (2005). Sovereign spread in emerging markets: A principal component analysis. Documentos de Trabajo (Banco Central de Chile), (333), 1.
11. Silvestre, J. L., villar, O. Dependance and financial contagion during the great recession.
12. Martinez, C., & Ramirez, M. (2011). International propagation of shocks: an evaluation of contagion effects for some Latin American countries. Macroeconomics and Finance in Emerging Market Economies, 4(2), 213-233.
13. Mollah, S., Zafirov, G., & Quoreshi, S. (2011, September). Financial Market Contagion During Global Financial Crisis. In Midwest Finance Association 2012 Annual Meetings Paper.
14. Meric, G., Lentz, C., Smeltz, W., & Meric, I. (2012). International evidence on market linkages after the 2008 stock market crash. The International Journal of Business and Finance Research, 6(4), 45-57.
15. Kritzman, M., Li, Y., Page, S., & Rigobon, R. (2010). Principal components as a measure of systemic risk. Available at SSRN 1582687.
16. Allen, F., & Gale, D. (2000). Financial contagion. Journal of political economy, 108(1), 1-33.
17. Avanzini, D., & Jara, A. (2013). A PCA approach to common risk exposures in the Chilean banking system (No. 707). Central Bank of Chile.
18. Corsetti, G., Pericoli, M., & Sbracia, M. (2005). ‘Some contagion, some interdependence’: More pitfalls in tests of financial contagion. Journal of International Money and Finance, 24(8), 1177-1199.
19. Jolliffe, I. T. (2002). Springer series in statistics. Principal component analysis, 29, 912.
20. Pericoli, M., & Sbracia, M. (2003). A primer on financial contagion. Journal of economic surveys, 17(4), 571-608.
21. Joy, O. M., Panton, D. B., Reilly, F. K., & Martin, S.A. (1976). Comovements of major international equity markets. Financial Review, 11(1), 1-20.
22. Lau, S. T., & McInish, T. H. (1993). Comovements of international equity returns: a comparison of the pre-and post- October 19, 1987, periods. Global Finance Journal, 4(1),1-19.
23. Shik Lee, H. (2004). International transmission of stockmarket movements: a wavelet analysis. Applied Economics Letters, 11(3), 197-201.
24. Meric, G., Leal, R. P., Ratner, M., & Meric, I. (2001). Co- movements of US and Latin American equity markets before and after the 1987 crash. International Review of Financial Analysis, 10(3), 219-235.
25. Meric, I., Coopersmith, L. W., Wise, D., & Meric, G. (2002). Major stock market linkages in the 2000-2001 bear market. Journal of Investing, 11(4), 55-62.
26. Shlens, J. (2014). A tutorial on principal component analysis.

Appendix 1: Correlation Matrix Before the Oil Crisis

Appendix 2 : correlation matrix during the oil crisis