WEAK FORM OF MARKET EFFICIENCY - EUROPEAN CAPITAL MARKET

Transcription

1 WEAK FORM OF MARKET EFFICIENCY - EUROPEAN CAPITAL MARKET REGEP HORAŢIU DAN PHD STUDENT AT FACULTY OF ECONOMICS AND BUSINESS ADMINISTRATION, WEST UNIVERSITY OF TIMISOARA, ROMANIA, horatiuregep@yahoo.com Abstract One of the most important implications of different levels of market efficiency is associated to the usability and the results of applying various methods of information processing. More specifically, it can be argued that the use of technical analysis is conditioned by the degree of market efficiency in terms of accuracy of trading signals generated by specific indicators. If financial market has a high degree of efficiency then the indicators are not able to generate systematic higher results than the market average. Conversely, if the market efficiency is low investors using technical analysis tools are able to build and manage portfolios of financial assets where the return/ risk ratio is higher compared to other portfolios were investors do not use a systematic treatment of the available information. The objective of this work lies in the description of the potential influence of financial and non financial information might have on financial markets, thus influencing the use of technical analysis tools in building investment portfolios. Thus, we analyzed the market efficiency in week form (autocorrelation test, run test, variance ratio test) in emerging capital market from Eastern Europe (WIG) compared to developed capital markets in Western Europe (Euro Stoxx). In this paper we focused on a individual analysis of the company stocks from those indexes divided according to the sector in which the companies operate (Consumer goods; Goods, Industry, Energy; Financial; Telecommunication). The analysis is done on an hourly frequency. Key words: market efficiency, autocorrelation test, run test, variance ratio test, capital market JEL Classification: G0, G4. Introduction One of the most important implications of different levels of information efficiency is associated to the possibility of using and the results of applying various methods of information processing. More concretely, one can argue that information can influence the effectiveness of investment strategies of capital market investors. The objective of this work lies in the description of the potential influence of financial and non financial information might have on financial markets, thus influencing the use of technical analysis tools in building investment portfolios. If the financial market shows a high degree of informational efficiency, these specific indicators related to technical analysis are not able to generate in a systematic manner superior results than the market average. Conversely, if the informational efficiency is low, investors using technical analysis tools have the possibility to build and manage portfolios of financial assets that, in the light of return/ risk ratio, are superior to portfolios of investors that do not use a systematic use of information available.. Market Efficiency in Eastern Europe - Literature In the last decade Eastern European countries have experienced rapid growth of capital markets. This increase was determined by their integration into the European Union and foreign direct investment flows (Cartensen and Touba []; Wolff [3]). The development of capital markets increases investment choices of investors to diversify portfolios. The informational efficiency in the allocation of financial assets becomes important from a global perspective. Informational efficiency has important implications for investors seeking to identify appropriate financial assets to invest in capital markets. If the market is informational efficient, there cannot be found undervalued assets to achieve high profits, these being valued at a price related to the risk exposure. However, if the market is inefficient, 0

2 investors could identify undervalued to financial assets and increase return. Efficient market hypothesis has two functions, first as a theoretical model of the functioning of the capital market, and second as a tool to attract savings on the capital market (Will []). Understanding the capital markets in developing countries is becoming increasingly important due to globalization and the free movement of investments. Studies of efficiency on the capital markets among developed markets are numerous, and these markets show a low level in terms of the weak form of efficiency, and a high level in terms of semi-strong form of efficiency. In contrast, studies of informational efficiency of emerging capital markets are few and contradictory results (Gupta and Basu [8]). In these cases the contribution of capital market in the investment development process is much smaller, which generates much weaker results with different restrictions and controls (Gupta [7]). Compared to other geographical areas, few studies have tested the information efficiency of capital markets in Central and Eastern Europe. Through analyzing the Polish stock market index WIG on a daily basis for the period , Nivet[9] showed that market returns do not follow the "random walk" hypothesis, thus Polish capital market does not show an informational efficiency in its weak form. By using VR and ADF test, Chun [] shows that the capital market in Hungary shows a random walk". The behavior of this market is more similar to those from Western Europe than those from Eastern Europe. The main reason is the high presence of foreign investors. Contrary to these results, Gilmore and McManus [5] show by using VR and Multiple Variance Ratio test that in the case of daily returns on the capital market in Hungary, Poland and the Czech Republic for the period July September 000, the "random walk" hypothesis does not stand. These results could be due to capital market imperfections that interfere with fast information processing and rare trading days. The main conclusion of these studies is that the capital market in Poland and Hungary are not yet informational efficient in a semi-strong form. Rockinger and Urga [0] examine the informational efficiency for several Central and Eastern European capital markets for the period April June 999 on a daily basis. They found that the Hungarian capital market shows information efficiency in its weak form, while the capital market in the Czech Republic and Poland are not informational efficient, but tend toward informational efficiency. Worthington and Higgs [5] tested the "random walk " hypothesis on capital markets in the Czech Republic, Hungary, Poland and Russia using Unit-Root test, VR and Multiple Variance Ratio. Following the analysis, only Hungary confirmed the "random walk" hypothesis, meaning a weak form of information efficiency. Guidi [6] analyzed a number of capital markets in Central and Eastern Europe during By using autocorrelation test, "run" test, and VR, he has identified that capital markets do not follow the "random walk" hypothesis. 3. Methodology To evaluate the weak form of informational efficiency, there have been developed three types of tests (autocorrelation test, "run" test, variance ratio test): - Tests that examine to what extent abnormal returns are independent of the quantity of available information at a precise moment, or earlier; - Tests that assess the manner through which certain investment strategies can provide above-average profits, taking into account the transaction costs and systematic risk of financial instruments studied; - Tests which verify if market prices are always equal to the accounting value. They use historical data to determine the accounting value, or variations of it. Consequently the focus is on the extent to which current price variation is similar to the fundamental indicators variation. Autocorrelation is a statistical tool that is used to measure the dependence of a variable on its previous values. Autocorrelation measures the relationship between the return of a financial instrument in a current period and that in a prior period. n t= k t t k ρ k = n () ( r t= r)( r ( r r) r) ρ k - serial correlation coefficient of financial instrument returns for a temporary lag k n - number of observations r t - return of the financial instrument during period t r t-k - return of the financial instrument during period t-k _ r - average return of financial instruments k - temporary gap 03

3 The test is intended to determine whether the series correlation coefficient is different from zero. Statistically, weak efficiency market hypothesis should be rejected if the financial instruments returns are serial correlated ( ρ is significantly different from zero). To analyze if a group of autocorrelations are simultaneously equal to 0, one takes into consideration the null hypothesis and uses Ljung-Box Q test. If Q-statistic is significantly different from zero it means that there is a level of autocorrelation, which would assume that the null hypothesis, stating that returns are independent, does not apply. "Run" test is a non-parametric test of designed to assess whether successive price changes are independent. There is a possibility of financial instruments prices follow trends that other rules cannot detect, indicating that variations are random during most of the time, but occasionally become auto-correlated time for variable periods of time. The run test analyzes exactly this possibility, and thus is used in a series of numbers, each time changes between successive observations change sign. The method of this test relates to a temporary independence between series. Thus, the hypothesis is tested by observing the number of successive sequences of price changes with the same sign, positive, zero or negative (Campbell et al., 997). Each change in return is compared to the average return. Thus, there is a positive change when the return is higher than average return, a negative change implies that the return is lower than average return, and zero is recorded when the return is equal to the average return (Worthington and Higgs, [4]). To perform the run test there will be a comparison between the observed current number of runs R and estimated number of runs m using the following equation: = 3 N( N + ) n i i m = N () N - number of observations i sign of plus, minus or no change n i - total number of changes for each sign category For a larger number of observations (N> 30) the estimated number of runs m is approximately normally distributed with a standard deviation of σ m rallies calculated as follows: m = ni ni + N( N + N i= i= i= N ( N ) n 3 N 3 / σ (3) Then in order to run the run test, Z-statistic is determined by the formula: Z R m ± 0,5 σ = ~N(0,) (4) m R - is the current number of runs As pointed out by Abraham (00), when the current number of runs exceeds or falls below the estimated number of runs, Z is positive or negative, respectively. A negative value of Z indicates a positive correlation in the series, and a positive value of Z indicates a negative correlation in the series. A positive correlation implies that there is a positive dependence of the financial instruments prices, indicating therefore a contradiction of the random walk theory. Variance ratio test (Lo and MacKinlay, 988) arose as a result of the fact that the variance of "random walk" increases linearly with time. k σ ( q ) VR q = (5) σ σ (q) - Unbiased estimator of / q of the variance of q difference 04

4 σ () - Variance of the first difference Assuming homoscedasticity, the first statistic test of z (q) is expressed as follows: z( q) VR( q) = ~ N (0.) (6) v( q) where: V (q) = [ ( q -) (q -)] / 3q (nq) (7) The second statistic test z*(q) is built on the same hypothesis of homoscedasticity as follows: z *( q) where: VR( q) = ~ N (0.) (8) v *( q) q ( q k) v*( q) = φ ( k) (9) k = q and : nq ( x x ˆ) µ ( x x t= k + k ˆ) µ φ ( k) = (0) t t t k t nq [ ( x ] t xt ˆ) µ t= Both the statistic test z(q) and the z*(q) test the null hypothesis that VR (q) = (selected index has a random evolution). If the random walk hypothesis is rejected and the VR (q)>, then the returns are positively serial correlated. As pointed out by Urrutia (995), for emerging markets, a serial positive correlation of returns describes a market growth. When the random walk hypothesis is rejected and VR (q) < then returns are negatively serial correlated, and this process is described as a process of "return to average". The methodology chosen consists in applying statistical tests above mentioned both with hourly and daily frequency to identify whether this factor may have an impact on informational efficiency. In the first phase, autocorrelation test is performed to identify whether consecutive returns are independent of each other. In the second phase there will be a non-parametric test (run test) also to identify whether changes in returns are independent of each other. In the last phase VR test will be conducted to verify the "random walk" hypothesis. 4. International Data and Results There has been conducted a study of informational efficiency in the weak form on capital markets in Eastern Europe (WIG) compared to developed capital markets in Western Europe (Euro Stoxx). In this paper we focused on a detailed analysis on high frequency data (hourly data) for each company that is part from the mentioned indexes. The period considered for the for the hourly frequency it is October 0 - April 03. All information in this study were obtained from Teletrader []. The information used are the closing prices for shares of companies that form the structure of DJI, Euro Stoxx and WIG indexes. To perform tests on informational efficiency hypothesis in the weak form, information was processed and returns were calculated with the following formula: t Pt = ln P R () t R t - return P t - financial instrument price at time t All the below tests (autocorrelation, run, variance ratio) are done for 6 companies that are part from Euro Stoxx Index. 05

5 The first approach for the analysis of informational efficiency in weak form is autocorrelation test on different lags (,, 0). Autocorrelation (Table) for the majority of stocks for all lags has an associated probability close to. This situation implies acceptance of the null hypothesis of random walk. However there are a number of stocks which for low lags show a corresponding probability close to, but as the lags are higher, the corresponding probability is decreasing close to 0. This means that for these companies, the dynamic of returns of shares issued is a "short term memory" process. Also, there are a number of stocks of companies, where the corresponding probability is close to including for higher lags. This implies that the dynamic of returns of shares issued is a "long term memory" process. So, in the case autocorrelation test we can mention that we have an informational efficiency in weak form. Since the data on hourly frequency does not follow a normal distribution, according to Jarque-Bera test, it is necessary to carry out the run test in order to verify the result obtained by the autocorrelation test. Table show results of the run test and it can be observed that most companies' stocks have a positive z- statistic which implies that it does not exist a positive serial correlation between returns, which further implies there is an independence of financial instruments prices indicating therefore the random walk hypothesis is accepted. The only stocks for which the random walk hypothesis is not verified and have positive serial correlation of returns are BAY, DAI, BAS and DEB. By using VR on daily data for stocks of companies from within the Euro Stoxx structure of the capital market in Western Europe returns varying results depending on how the properties of the analyzed data were taken into account. In the first stage, using a simple form of random walk all results strongly reject the null hypothesis. Both the Max z test and the Wald test showed probability of zero when analyzing all shares of the companies and likewise for the index. These results show that regardless of sector, the market analyzed is not efficient. Nevertheless, when innovation was incorporated in the analysis, the results have changed. Thus, in this case, the null hypothesis is not rejected in all cases. This result suggests that the shocks that disturb the returns dynamics cannot be described as a random walk process type. According to the results in table 4, we can see that in financial sector, shares of DEB company do not reject the null hypothesis, and as for telecommunications sector companies we have the same result for DET and SAP shares. VR test does not reject the null hypothesis. From the analysis of the sectors we can see that in very few cases the null hypothesis is rejected, implying that regardless the sector, the market is considered efficient. From the analysis performed on shares of companies in the structure of ESX, mixed results were recorded depending in the tests applied. Thus, in case of autocorrelation the random walk null hypothesis is accepted. For the run test, the null hypothesis is accepted. For VR, when the simple variant of random walk is considered, the null hypothesis is rejected. When innovations are introduced, the null hypothesis is accepted. All the below tests (autocorrelation, run, variance ratio) are done for 3 companies that are part from WIG 0 Index. The first approach for the analysis of informational efficiency in weak form is autocorrelation test on different lags (,,... 0). There is shown in table that autocorrelation for the majority of stocks for all lags have an associated probability close to zero, which implies rejection of the null hypothesis of random walk. This implies that the relationship between current returns and its previous value is significant. Regardless of the results obtained by the autocorrelation test in order to verify the result we will go forward we the run test Tables 3 show results of the run test and it can be observed that most companies' stocks have a positive z- statistic which implies that it does not exist a positive serial correlation between returns, which further implies there is an independence of financial instruments prices indicating therefore the random walk hypothesis is accepted. Only for company LOT shares, the random walk hypothesis is not confirmed. The use of VR for shares of companies from within the WIG structure of the capital market in Eastern Europe returns varying results depending on how the properties of the analyzed data were taken into account. In the first stage, using a simple form of random walk, all results strongly rejected the null hypothesis. Both the Max z test and the Wald test showed probability of zero when analyzing all shares of the companies and likewise for the index. These results show that regardless of sector, the analyzed market is not efficient. Nevertheless, when innovation was incorporated in the analysis, the results have changed. Thus, in this case, the null hypothesis is not rejected in all cases. This result suggests that the shocks that disturb the returns dynamics cannot be described as a random walk process type. According to the results shown in table 5 we can see that in the goods sector, industry and energy, there are 3 companies whose shares do not reject the null hypothesis: KGH, LOT and PKN. In the financial sector, the shares of PKO do not reject the null hypothesis. Analyzing the hourly frequency for WIG index by using the VAR test, the null hypothesis is not rejected. The analysis of all our sectors shows that in many cases the null hypothesis is rejected, implying that the market is not considered weak form efficient in terms of information. The analysis performed on various companies which are part of the WIG structure recorded mixed results depending on the type of tests applied. Thus, considering autocorrelation the null hypothesis is rejected. The run test accepts the null hypothesis. VR rejects the random walk null hypothesis. When innovation is introduced in the equation, the null hypothesis is also rejected. 06

6 5. Final Consideration One of the most important implications of different levels of market efficiency is associated to the usability and the results of applying various methods of information processing. More specifically, it can be argued that the use of technical analysis is conditioned by the degree of market efficiency in terms of accuracy of trading signals generated by specific indicators. If financial market has a high degree of efficiency then the indicators are not able to generate systematic higher results than the market average. Conversely, if the market efficiency is low investors using technical analysis tools are able to build and manage portfolios of financial assets where the return/ risk ratio is higher compared to other portfolios were investors do not use a systematic treatment of the available information. Advanced analysis illustrates that we obtained different result in weak form of market efficiency depending on the frequency, region, sector that were considered. Consequently, we consider that there are many opportunities for the use of technical analysis in the construction of diversified portfolios which provide a better management of the various types of subsequent risks. Tables Table Autocorrelation of ESX and WIG Index hourly frequency ESX WIG Lags AC Q-Stat Prob AC Q-Stat Prob Retunr<= Average Table Run test ESX structure hourly frequency return> average Obs No of runs runs estimated z- statistic ESX ,74 -,4 0,5 Sector - consumer goods BAY ,00-0,36 0,7 DAI ,5-0,09 0,93 VOL ,5 0,3 0,90 Sector goods, industry, energy BAS ,99-0,50 0,6 Prob> z 07

7 Retunr<= return> No of runs z- Prob> Average average Obs runs estimated statistic z BAR ,63,8 0,4 EON ,56 0,5 0,80 EUR ,4,00 0,3 RWE ,74,90 0,06 SIE ,83,39 0,0 TOT ,87 0,0 0,99 Sector financial ALZ ,98 0,58 0,56 DEB ,49-0, 0,9 MUE ,98,0 0,3 Sector telecommunications DET ,34 3,03 0,00 SAP ,98,09 0,04 TEL ,94,46 0,0 Table 3 Run test WIG structure hourly frequency Retunr<= Average return> average Obs No of runs runs estimated z- statistic Prob> z WIG ,46,0 0,04 Sector goods, industry, energy BOR ,93 5,9 0,00 KGH ,4 0,40 0,69 LOT ,6-0,59 0,56 PGN ,08,79 0,0 PKN ,78 0,6 0,53 Sector financial BRE ,55 3,8 0,00 GTC , 0,73 0,46 HAN ,00 3,0 0,00 PEK ,7,08 0,04 PKO ,87 0,80 0,4 Sector telecommunications ASS ,80,04 0,04 TPS ,04,7 0,09 TVN ,39,40 0,0 08

8 Table 4 Variance ratio test ESX structure hourly frequency Random walk Random walk innovations Wald Wald Max z (Chi-Square) Max z (Chi-Square) ESX 3, 74,75 0,9,43 Prob 0,00 0,00 0,67 0,84 Sector consumer goods BAY,54 58,7 0,9,73 Prob 0,00 0,00 0,66 0,6 DAI 3,74 88,99,3,0 Prob 0,00 0,00 0,43 0,75 VOL,83 64,75 0,9,9 Prob 0,00 0,00 0,66 0,7 Sector goods, industry, energy BAS,6 5,68 0,63 7,93 Prob 0,00 0,00 0,85 0,08 BAR,89 66,09,30 5,45 Prob 0,00 0,00 0,05 0,6 EON 3,99 98,7,0 5,0 Prob 0,00 0,00 0,48 0,9 EUR 3,75 90,0,5 7,5 Prob 0,00 0,00 0,8 0,09 RWE 3,94 94,8,03 4,64 Prob 0,00 0,00 0,0 0,33 SIE 4,7 04,88,64 7, Prob 0,00 0,00 0,0 0,4 TOT,99 69,84,0,40 Prob 0,00 0,00 0,60 0,84 Sector financial ALZ,77 63,5,47 5,97 Prob 0,00 0,00 0,30 0, DEB,7 38,84,97 4,70 Prob 0,00 0,00 0,0 0,0 MUE 3,03 69,84,56 4,3 Prob 0,00 0,00 0,7 0,38 Sector telecommunications DET 5,7 54,09 3,4 5,43 SAP 3,8 9,78,99 0,5 Prob 0,00 0,00 0,0 0,04 TEL 3,5 8,87,7 3, Prob 0,00 0,00 0,50 0,56 09

9 Table 5 Variance ratio test WIG structure hourly frequency Random walk Random walk innovations Max z Wald (Chi-Square) Max z Wald (Chi-Square) WIG 5,68 46,,7,88 Prob 0,00 0,00 0,43 0,59 Sector goods, industry, energy BOR 9,4 387,64,66 36,83 KGH 5,64 44,8 0,69 3,5 Prob 0,00 0,00 0,8 0,54 LOT 6,08 60,94,83 6,38 Prob 0,00 0,00 0,4 0,7 PGN 6,6 78,58 4,00 8,93 PKN 6, 60,63,0,6 Prob 0,00 0,00 0,48 0,67 Sector financial BRE 6,38 68,9 5,6 33,98 GTC 6,60 76,76,8 0,56 Prob 0,00 0,00 0,0 0,03 HAN 8,09 336,6 6,7 38,98 PEK 5,85 5,33 3,69 5,9 PKO 5,03 6,38,7 3,73 Prob 0,00 0,00 0, 0,47 Sector telecommunications ASS 6,4 70,79 3,79 5,30 TPS 5,77 50,06,8 8,8 Prob 0,00 0,00 0,05 0,07 TVN 6,9 89,4 3,49 3,35 Prob 0,00 0,00 0,00 0,0 Acknowledgments This work was cofinaced from the European Social Fund through Sectoral Operational Programme Human Resources Development , project number POSDRU/59/.5/S/40863, Competitive Researchers in Europe in the Field of Humanities and Socio-Economic Sciences. A Multi-regional Research Network. 0

10 References [] Cartensen, K., Toubal, F. (004), Foreign Direct Investment in Central and Eastern Europe Countries: A Dynamic Panel Analysis, Journal of Comparative Economics, 3: 3-; [] Chun, R.M. (000), Compensation Vouchers and Equity Markets: Evidence from Hungary, Journal of Banking and Finance, 4: 55-78; [3] Gábor, K., Tamás, S. (04): What are the Differences Between the Currencies of Foreign Exchange Loans?. Public Finance Quarterly, vol. 59, issue, p [4] Gábor, K., Andreász, K (0): The Impact of the Crisis on the Monetary Autonomy of Central and Eastern European Countries. Public Finance Quarterly, vol. 57, issue, p [5] Gilmore, G.C., McManus, M.G (003), Random-walk and efficiency tests of Central European equity markets, Journal of Managerial Finance, 9 (4): 4-6; [6] Guidi, F. (0), Weak-form Market Efficiency and Calendar Anomalies for Eastern Europe Equity Markets, Journal of Emerging Market Finance, Institute for Financial Management and Research, 0(3), ; [7] Gupta, R. (006), Emerging Market Diversification: Are Correlations Changing Over Time?, In International Academy of Business and Public Administration Disciplines (IABPAD)Conference, Orlando; [8] Gupta, R., Basu, P.K. (007), Weak Form Efficiency in Indian Stock Markets, International Business & Economics Research Journal, 6 (3): 57-64; [9] Nivet, J.F. (997), Stock Markets in Transition: The Warsaw Experiment, Economics of Transition, 5: 7-83; [0] Rockinger, M., Urga, G. (000), The Evolution of Stock Markets in Transition Economies, Journal of Comparative Economics, 8: 456-7; [] Teletrader (03), Teletrader Proffessional, available at [] Will, M. (006), The Role of Performances in an Accounting Scandal: An insider s Perspective on How Things Went Wrong, Proceedings of the 5th Global Conference on Business & Economics, Cambridge University, UK, 5; [3] Wolff, G.B. (006), Foreign direct investment in the enlarged EU: do taxes matter and to what extent, Deutsche Bundesbank, Discussion Paper Series, Economic Studies, 3, -60; [4] Worthington, A.C., Higgs, H. (004), Random walks and market efficiency in European equity markets, Global Journal of Finance and Economics, : 59-78; [5] Worthington, A.C., Higgs H. (008), Tests of random walks and market efficiency in Latin America Stock Markets: an empirical notes, working paper No 57, School of Economics and Finance, Queensland University of Technology;