AN INVESTIGATION INTO THE EFFECT OF FUNDAMENTAL ANALYSIS ON STOCKS ABNORMAL RETURN IN THE COMPANIES LISTED ON TEHRAN STOCK EXCHANGE Mohammad Ashoub *, Abdolhamid hoshmand **13 * Department Of Accounting, Sarvestan Branch, Islamic Azad University, Sarvestan, Iran ** Department Of Accounting, management and economic college, Shiraz Branch, Islamic Azad University, Shiraz, Iran Abstract This study evaluates the ability of ratios in explaining the value of stocks, returns, incomes, and rate of accounting return. It aims to study the extent to which other accounting variables, in addition to income, can forecast future status. In order to conduct this study, the companies listed on Tehran Stock Exchange between 2006-2009, for which the end of financial year is March, have been used. This study is correlation and its methodology is post-event (using past data). The hypotheses are tested using multi-variant regression models based on each year s data. The results obtained from this study indicate that the total return and excess return on the market are not predictable based on accounting entities. The results also indicate that using models that are based on accounting entities to forecast excess return on the market is not significantly different from a simple model based on net incomes. Keywords: Fundamental analysis, Stocks abnormal return, Total return 1. Introduction The science of accounting aims to prepare valuable data to forecast different events so to facilitate decisionmaking. Economic decisions are in need of reliable information and the system of accounting has been designed so to meet this need. Founding the structure of the theoretical framework of accounting on the benefit of decision-making is another evidence. To this aim, accounting system has a lot of audience and in order to make preparation and provision of data more coherent, investors and financial analysts have been regarded as the most significant users. Financial statements as annual financial statements and performance of companies include valuable information for users. In this regard, financial ratios, as the summary of financial statements, provide useful information on the internal state of companies. In actual fact, one of the consequences of evolution of accounting has been using the financial ratios to analyze financial statements. Literature on forecasting the return based on ratio focuses on the power of predicting future income using financial ratios. Experimental evidence is compatible with the ability of ratios in predicting the growth of income. With regard to income forecast, research is categorized into a number of main areas such as forecasting through the use of time series of annual incomes, evaluation of ability of previous seasonal incomes in forecasting income, research on the role of components of income in forecasting, and evaluation of the forecasting ability of financial items or variables related to industry as well as major economic variables (Jahankhahi and Abdeh Tabrizi, 1993:13). 2. Theoretical Background and Review of Literature Accounting experts are particularly interested in capital markets. This is because observing the market reactions to accounting information is a tool to test accounting theories. This was not possible in the past, the unsolvable mystery on the emphasis on the usefulness of data for each and every investor is mainly an individual decisionmaking. Even if it is possible to measure individuals mental interest, the accumulation of these measures for a particular group of investors is difficult. Therefore, the real test that what type of accounting data is more useful 13 Corresponding Author COPY RIGHT 2012 Institute of Interdisciplinary Business Research 493
for each and every investor is highly difficult, if not impossible. It seems that market test has been a solution for this problem and has enabled researchers to emphasize the effect of information on investors as one of the main principles in the market. Different studies have been done in recent years to determine how the above-mentioned objectives could be achieved. The reason behind these researchers is to test this hypothesis: does the stock market act logically with regard to acquiring and processing input data? Is the data immediately and without any particular trend reflected on the price of equities? (Jahankhani and Abdeh Tabrizi, 1993: 13). One of the first findings in these researches was that changes in stock price are not correlated in the capital market. Fama (1970) stated that lack of correlation is due to the efficiency of the capital market. In other words, the price of equities in the market reflects all accessible data without any particular trend and no related data is ignored by the market (p.416). The studies done in this area mainly use such ratios as coefficient and ratio of price to book value as the indices for future income growth. They also use fundamental variables such as rights of stockholders, net income, dividend and stock price to explain and to forecast changes in incomes. Freeman et al. (1982) investigated the role of book return in forecasting changes in incomes. The results of their study revealed that the rate of book return follows a process of reversing to average and changes in book return are strongly related with the changes in incomes. They demonstrated that the likelihood of an increase in incomes is not independent of the predictor of book return rate (p. 640). Ou and Penman (1989) conducted a series of strong academic researches with regard to forecasting incomes based on multi-variable analysis of financial ratios. They used statistical methods to reduce the large number of financial ratios to an efficient group to predict the income. In their selected samples, they demonstrated that the forecast model using the ratios acts better than the time series models of annual incomes. They also showed the panel data of individual ratios leads to obtaining a more accurate forecast of future profits (p.296). Penman (1996) extracted the information from the price to forecast future profitability. He developed techniques that combined the information in the ratios of price to profitability and ratios of price to book value in order to forecast future profitability or stock return (p.254). in addition, Abarbanell and Bushee (1997) investigated the relationship between fundamental variables and changes in future incomes. The results revealed that variables of inventory, margin of net income, effective tax rate, quality of income, and workforce were meaningfully related with the income of next year. Belkaoui (1997) investigated the relationship between financial ratios and stock value. Based on the results, they concluded that current ratio, income before tax deduction, long-term liability to rights of stockholders, and the total price of sold goods had a reverse correlation and variables of dividend of each stock, income before tax deduction to rights of stockholders, and net income to sales are directly correlated with return (p.194). Garood and Rees (1999) used the four fundamental variables of stockholders rights, net income, dividend, and stock price to explain and forecast changes in incomes. The variables that were compared with rights of stockholders showed considerable potential to explain the changes in the net incomes of next two years. Penman and Zhang (2000) conducted another study in the area of forecast. These researchers studied the interactions between changes in growth and methods of conservative accounting such as costs of R&D and marketing. The results of their studies showed that companies that declare the abnormal changes in their R&D and marketing costs in their Lifo savings experience a return to initial state in their return on net assets. Piotroski (2000) used a strategy of fundamental analysis for a portfolio of companies with a book value of high market value and showed how the distribution of the return obtained by an investor is increased. In addition, the choice of a value stock is a unique advantage for investigating the ability of fundamental analysis in choosing companies with high returns. This is due to the fact that value stocks usually have little analytic support and rarely attract investors. Therefore, financial statements are the most reliable data in these companies (p. 1). The study by Beaver and Ryan (2000) is another example of studies about forecast of incomes based on ratio. These two researchers separated the components of bias and lag in the ratios of price to book value in order to forecast future book return. They predicted that there is a reverse correlation between lag and future stock return. In other words, book value with high price has lower profitability growth. The time horizon of forecast, according to this logic, depends on the lag time or the speed of reflection of economic losses or profits on the book value. Experimental data is consistent with this interpretation (p.127). Beneish et al. (2001) used the two-stage approach of analysis of financial statements. In the first stage, they used the market-based signals in order to differentiate between companies with high and low performance. In the second stage, they used fundamental factors to distinguish between winners and losers in the companies with threshold performance of the first stage. Results of the study are indicative of the significance of fundamental analysis (p.187). a recent study done by Elleuch (2009) revealed that fundamental signals have a COPY RIGHT 2012 Institute of Interdisciplinary Business Research 494
positive and meaningful relationship with future performance, in a way that successful portfolios with high fundamental scores have better performance in comparison to unsuccessful portfolios. 3. Methodology 3.1. Operational Statement of Hypotheses of the study The study s hypotheses are as follows: H1: Excess return on the market is not predictable in models based on accounting entities. Statistical statement of this hypothesis is as follows: R R PTE i 0 i 8 ij Signals Excess return is equal to the result of the difference in realized return and expected return in the market i model. PTE i Annual change in income before tax deduction. j 1 ij u ij Signals ij Financial variables defined on the basis of the definitions in the variables section of the research. The first null hypothesis is as follows: H... 0 o1 1 2 j H2: forecasting excess return on the market using models based on accounting entities is not significantly different from a simple model based on net income. Statistical statement of this hypothesis is as follows: R E i 1 i Ri Excess return is equal to the result of the difference in realized return and expected return in the market model. E i Annual change in each stock income. u ij H3: total return is not predictable in models based on accounting entities. Statistical statement of this hypothesis is as follows: TR PTE 0 i 8 it Signals j 1 it u it TR : total return equal to the result of the difference between realized return and expected return in market model PTE i annual changes in incomes before tax deduction. Signals it : financial and non-financial variables defined in the variables section of the study. Null hypothesis 3 is as follows: H... 0 o1 1 2 j COPY RIGHT 2012 Institute of Interdisciplinary Business Research 495
3.2. Scope of the Study This study is correlation and its methodology is post-event. The time period of the study includes 2006 to 2009. In order to select the case studies, the following issues were taken into consideration: 1. The end of the financial year of companies was 20th of March. 2. The companies activities were not investment. 3. The company was continuously active in the period between 2006 and 2009. 4. The complete data of the company was available. In experimental studies, the data can principally be time series, in other words, they could be in the frame of one or a number of particular variables in a period of time. They could also be cross-section, i.e. they are calculated and collected in a particular cross-section of time for instance year. In studies like the present study, the data are usually obtained by combining two series of cross-section data and their time series. In this case, two types of data design is possible. In the first type, the data of one unit of cross-section for year are put together and then this is repeated for the second cross-section and the rest. This type of data design is known as pooling data. In the second type of data design, the data of cross-section units for each year could be put together in a way that the same trend is repeated for the next years. This type of data design is known as panel data. The present study uses pooling data. In order to test the first hypothesis, the following regression equation is fitted: R PTE ij Signals it 0 it 9 j 1 it u it R v, lneemv ivfsevgen evt ffeev ea ett ni it is the excess return of stock I, while calculation of the total return is began after four months into the financial year. The excess return is the result of the difference between realized return and expected return of the market model. In order to calculate the expected return of the market model, and PTE coefficients are calculated. it is annual change in profit before tax deduction. Other independent variables second statement of the above equation- include above fundamental signals in the Signals variables section of the research. In this equation, it indicates independent variables. In the next step, in order to test the second hypothesis, another regression equation will be fitted: R E it 0 it E v, ivfs tvesigf ni it is annual changes in income of each stock divided by the stock price at the beginning of the year. This statement is used as an index for equation 1 and the different in r2 coefficient will be the basis of judgment on performance of forecast model. u it 3.3. Definition of variables Nine fundamental variables have been used in this study in order to measure 4 areas of financial positions of companies: 1) Profitability variable: a) ROA b) ΔROA c) ROE 2) Operational efficiency: a) Δ TURN COPY RIGHT 2012 Institute of Interdisciplinary Business Research 496
b) Δ INVT c) Δ SLINV 3) Leverage and dividend: a) Δ LEVG b) Δ DIVEQ 4) Changes in profit before tax deduction: a) ΔPTE 4. Evaluation and Analysis of the model In order to assess the normality of the variables, non-parametric Kolmogorov-Smirnov Test was conducted and it was observed that sig / 53 0/ 05. Therefore, H1 does not have normal distribution. 4.1 Statistical test of the hypotheses The relationship between independent and dependent variables in this study is based on multi-variant regression. These regression models are linear. For each hypothesis, models are tested in general, at the level of sample companies, and at the level of industry. The objective is to investigate the significance of t statistic at acceptable error level. This statistic explains the significance of all regression model. Then, using the r2 coefficient, it would be possible to investigate the strength of fitted regression model in explaining dependent variable. At the next step of the test, in order to increase the strength of the model, coefficients of each and every independent variable was investigated with regard to dependent variable. The variables that are not meaningful at the first level of the test are eliminated from the model using elimination method. At this stage, the acceptable error level is 5%. At all stages of the test, if the Durian Watson statistic indicates a self-correlation between the remaining components (errors), the model will be modified using the self-reverse method and then the statistical test will be conducted again. When the result of the Kolmogorov-Smirnov Test shows abnormality in one of the variables, the logarithm of that variable in used in the fitted regression model. In these models, the following acronyms are used to show independent and dependent variables. DIVEQ: v e eeievenm id LEVQ: changes in capital structure of company ROA: voperational income divided by total assets PTE: visgiaetv ivincomev eeeiif niveetn evfgbv SLINV: changes in ratio of sales to inventoriesv INVT: changes in inventories ROA: changes in ratio of operational income to total assetsv RIv: excess return TURN: change in ratio of turnover of operational assetsvv TR: fnfgmv efi i vv v 4.2. e iatvtn vfsevt tfvsdenfset tn i H1: Excess return on the market is not predictable in models based on accounting entities. The correlation between variables has been demonstrated in Table 1. prob index in the tables indicates that level of significance of the correlation between the variables is under study. Prob in some variables at the significance level of lower than 0.05 rejects H0 (H0: there is no significant correlation between the two variables) and COPY RIGHT 2012 Institute of Interdisciplinary Business Research 497
Pearson correlation coefficient indicates the direction and strength of this correlation. Based on the findings of Table 2., it could be said that the variables that enter the model include: DIVEQ; DLEVG; DINVT; ROA As it could be seen, the correlation coefficient of the first model is 0.105. f-statistic confirms H0 (Ho= regression is meaningful) at the significance level of lower than 0.5. In order to provide a good model, errors need to be independent. Durbin-Watson test in the above table is 0.933 and in the significance level of 0.5>D- W<2.5, indicating that errors are independent. In addition, Hausman test is at level of 0.2386 and Pagan test at the level of 0.0129 in the significance level of lower than 0.5 rejects H0 (H0: regression has stable effects), indicating that model has random effects. Test of remainders shows a regular distribution, showing a good model. The final model of H1 could be stated as: R = 13.1527034769-5.02683905699e-06*DIVEQ + 1.13751871632*DLEVG - 0.0726737386502*DINVT + 22.4907479251*ROA + [CX=R] 4.3. findings of testing the second hypothesis: H2: forecasting excess return on the market using models based on accounting entities is not significantly different from a simple model based on net income. The correlation between variables has been demonstrated in Table 3. prob index in the tables indicates that level of significance of the correlation between the variables is under study. Prob in some variables at the significance level of lower than 0.05 rejects H0 (H0: there is no significant correlation between the two variables) and Pearson correlation coefficient indicates the direction and strength of this correlation. There is a random effect. As it could be seen, the correlation coefficient of the second model is 0.172. f-statistic confirms H0 (Ho= regression is meaningful) at the significance level of lower than 0.5. In order to provide a good model, errors need to be independent. Durbin-Watson test in the above table is 2.227 and at the significance level of 1.5>D- W<2.5, indicating that errors are independent. In addition, Hausman test is at level of 0.9711 and Pagan test at the level of 0.0129 rejects H0 (H0: regression has stable effects) at the significance level of lower than 0.5, indicating that the model has random effects. Test of remainders shows a regular distribution, showing a good model. The final model of H1 could be stated as: R = 9.5552926372 + 0.0323089968557*DELTAE + [CX=R] 4. 4. Finding of testing the third hypothesis H3: total return is not predictable in models based on accounting entities. There is a random effect. Based on the findings of the previous table, the variables that enter the model include: DIVEQ DLEVG DTURN ROA As it could be seen, the correlation coefficient of the third model is 0.106. f-tatistic confirms H0 (Ho= regression is meaningful) at the significance level of lower than 0.5. In order to provide a good model, errors need to be independent. Durbin-Watson test in the above table is 1.733 and at the significance level of 1.5>D- W<2.5, indicating that errors are independent. In addition, Hausman test is at level of 0.2377 and Pagan test at the level of 0.0129 rejects H0 (H0: regression has stable effects) at the significance level of lower than 0.5, indicating that the model has random effects. Test of remainders shows a regular distribution, showing a good model. The final model of H1 could be stated as: TR = 13.1826233381-5.00737932733e-06*DIVEQ + 1.13635180716*DLEVG + 0.272301146019*DTURN + 22.499788958*ROA + [CX=R] COPY RIGHT 2012 Institute of Interdisciplinary Business Research 498
5. Results of the Study Results of testing the hypotheses are as follows: Results of testing the hypotheses are as follows: H1: Excess return on market is not predictable in models based on accountingventitiesv. Independentv variables of this hypothesis include:v dividend policy, change in thev capital structure of the company, change in ratio of sales to inventoriesv,change in inventories, change in ratio of operational assets, change in ratiovof net income to rights of stockholders, change in ratio of operationalvincome to total assets, ratio of operational income divided by total assetsv,and changes in earnings before tax deduction. Amongv these variables, the four variables of dividend policy, change in thev capital structure of the company, change in inventories, and ratio of operationalvincome to total assetvhave the ability of predicting excess return andvhave significant correlation with the dependent variable. Therefore, it couldvbe concluded that these variables have excess return in forecast models. H2: forecasting excess return on the market using models based on accounting entities is not significantly different from a simple model based on net income. Independent variables in this hypothesis include: last realized EPS. This variable has significant correlation, thus the ability to forecast excess return. H3: total return is not predictable in models based on accounting entities. Independent variables of this hypothesis include: dividend policy, change in capital structure of the company, change in ratio of sales to inventories, change in inventories, change in ratio of turnover of operational assets, change in ratio of net income to rights of stockholders, change in ratio of operational income to total assets, ratio of operational income divided by total assets, and changes in incomes before tax deduction. Among the variables of the above model, the four variables of dividend policy, change in the capital structure of company, change in ratio of turnover of operational asset, and change in ratio of operational income to total asset are able to predict the total return and remained in the model because of having a significant correlation with total return. 5.1. Conclusion: comparison of models In order to evaluate the ability to forecast total return and excess return by financial ratios effective in forecasting, two models were fitted. The aim of modeling was to identify the useful ratios in forecasting the abnormal return rate and total return. Investigating H1 and H3, it was observed that the three variables of dividend policy, change in capital structure of the company, and ratio of operational income to total asset in both hypothesis has a significant correlation with dependent variables of excess return and total return and also is better able to forecast in comparison to other variables. In addition, comparing R² in model 1 and 3, it could be seen that the forecast power of variables in model 1 is R²=0.1057 that is almost equal to variables in model 3, i.e. R²:0.1058. Comparing H1 and H3, it was observed that variables of H1 with R²=0.1057 are less powerful in forecast in comparison to variables of model 2 with R²=0.1715. Therefore, H2 is not rejected. Investigating the signs of the coefficients of variables in Tables 1 to 3 is indicative of the direction of the correlation between remaining independent variables in the model and the independent variable in question. 5.2. Recommendations of the study 1. Institutions in the capital market are recommended to create professional centers to conduct professional activities with regard to analysis and to integrate their professional activities. Lack of easy access and lack of regulated and valid forecasts by professional analysts are among the shortcomings that could be seen in the practical and scientific environment of Iran s capital market. If such centers are created, using forecasts and evaluations by analysts in scientific researches related to the capital market and proper reactions by analysts of the results of these researches could lead to more dynamism in the capital market. In such studies as the present study, use of forecasts by analysts about future income could strengthen the forecast models and lead to better results. COPY RIGHT 2012 Institute of Interdisciplinary Business Research 499
2. In the Iranian investment environment, determining the value and fundamental analysis have been ignored to a great extent. Investment institutions, minor investors, stockbrokers, and other players in the market are all influenced by the climate of the market and purchase and sell stocks without enough analysis and proper perspective. Unfortunately, purchase or sell queues play a major role in investors decision-makings. This factor alone has made the Iranian stock exchange one of the riskiest markets in the world. This shows that the shallow environment of analysis could be attributed to all components of the market, including companies and professional investment institutions. It is recommended that investors and particularly professional investment institutions pay more attention to scientific analysis and use methods of value-setting to determine the real value of equities. Since this study found a correlation between future income and financial ratios, it is recommended that analysts and investors pay more attention to fundamental analysis and particularly analysis of ratios in their assessments. COPY RIGHT 2012 Institute of Interdisciplinary Business Research 500
References Abarbanell, J., & Bushee B. (. Fundamental Analysis, Future Incomes and Stock Prices. Journal of Accounting Research,, Beaver, W., and Ryan S. (2000). Biases and lags in book value and their effects on the ability of the book-tomarket ratio to predict book rate of return on equity. Journal of accounting research, 38, 127-148. Belkaoui A. (. Value Relevance of Popular Financial Ratios. Advances in quantitative analysis of finance and accounting,5, Beneish, M. D., Lee, C.M., & Tarpley R. L. (2001) Contextual Financial Statement Analysis Through the Prediction of Extreme Returns. Review of Accounting Studies, 6, 165-189. Elleuch J. L. (2009). Fundamental Analysis Strategy and the Prediction of Stock Returns. International Research Journal of Finance and Economics, 30, 95-107. Fama E. F. (1970). Efficient Capital Market: A Review of Theory Empirical work. Journal of Finance, 25 (2), 383-417. Freeman, R.,Ohlson, J., & Penman S. (1982). Book rate of return and prediction of incomes changes: An empirical investigation. The Journal of Accounting Research, 20 (2), 639-653. Garrod, N., & Rees W. (1998). Forecasting incomes growth using fundamentals. Working paper, University of Glasgow. Jahankhahi, A., & Abdeh Tabrizi H. (1993). Capital market Efficiency Financial researches,1, 1-23. Ou, J. A., & Penman S. (. Financial Statement Analysis and the Prediction of Stock Returns. Journal of Accounting and Economics, 11, 6- Penman S. (. The Articulation of Price Incomes Ratios and Market to Book Ratios and the Evaluation of Growth. Journal of accounting research,,. Penman, S., & Zhang X. (. Accounting Conservatism, the Quality of Incomes, and Stock Returns. Working paper, Colombia university. Piotroski J. (2000). Value Investing: The Use of Historical Financial Statement Information to Separate Winners from Losers. Journal of Accounting Research, 38, 1-42. COPY RIGHT 2012 Institute of Interdisciplinary Business Research 501
Annexure Tables (No. 1 & 2) of testing the first hypothesis: eg geme C DIVEQ DLEVG DSLINV DINVT DTURN DROE DROA ROA DPTE ieeb 0/0002 0/019 0/0134 0/8022 0/0098 0/6919 0/56 0/8255 0/0285 0/9167 tfgf tf i -f 3/73613 2/569802 2/621436-0/250594 2/851226 0/396451 0/583175-0/220588 1/946751 0/104617 DS 3/519012 8/82 1/830469 0/052357 0/480565 0/681566 3/466647 32/18401 15/54570 1/88 Table 1: Correlation between variables tnett i eif 13/15270-5/3 1/137519-0/01310120-0/072674 0/270207 2/021664-7/099402 22/49075 1/97 0/0136 0/0129 f-tfgf tf i observation R² v 4/3251 8/698 f-tfgf tf i observation R² v Table 2 Correlation series Test of breusch pagan Tables (No. 3, 4, 5 & 6) of testing the second hypothesis: Variable C DIVEQ DLEVG DSLINV DINVT DTURN DROE DROA ROA DPTE index t-statistic SD 0 4/297477 5/021631 0/4201-0/807 2/38 0/6876 0/402 2/073 0/8075-0/2438 0/054163-0/6733-0/4219 0/5412 0/7018 0/3831 0/6913 0/6861 0/404 3/926 0/3518 0/932 37/328 0/1115-1/595 35/514 0/7186-0/3605 1/95 Table 3: Correlation between variables Coefficient 21/58034-1/92 0/834-0/0132-0/2286 0/2648 1/588 34/793-56/636-7/03 COPY RIGHT 2012 Institute of Interdisciplinary Business Research 502
b iee tfgf tf i -f DS tnett i eif eg geme 0/0101 2/585 3/697 9/555 C 0 5/275 0/0061 0/0323 DELTA E Table 4 ieeb tfgf tf i -f DS tnett i eif eg geme 0/0553 1/924 4/905 9/438 C 0/0113 2/550 0/013 0/0327 DELTA E Table 5 0/0001 0/0001 f-tfgf tf i observation ² v R 10/101 19/423 f-tfgf tf i observation R² v Table 6: Correlation Series Test-Pagan method Tables (No. 7, 8 & 9) of testing the third hypothesis: eg geme ieeb tfgf tf i -f DS tnett i eif C 0/0002 3/746 3/519 13/183 DIVEQ 0/0105 2/568 8/82-5/01 DLEVG 0/0098 2/621 1/831 1/136 DSLINV 0/7977-0/2564 0/052360-0/01342 DINVT 0/878-0/153 0/481-0/0736 DTURN 0/0197 2/399 0/681 0/272 DROE 0/560 0/583 3/466 2/0195 DROA 0/8284-0/2168 32/185-6/980 ROA 0/0183 1/947 15/546 22/499 DPTE 0/9186 0/102 1/88 1/92 Table 7 COPY RIGHT 2012 Institute of Interdisciplinary Business Research 503
eg geme C DIVEQ DLEVG DSLINV DINVT DTURN DROE DROA ROA DPTE ieeb 0 0/402 0/6878 0/8030 0/6714 0/699 0/686 0/3489 0/1109 0/7172 tfgf tf i -f 4/305-0/806 0/402-0/2495-0/424 0/386 0/404 0/9376 6647966 - -0/362 DS 5/022 2/38 2/073 0/054 0/541 0/691 3/926 37/33 35/51 1/95 tnett i eif 21/621-1/92 0/833-0/013519-0/229 0/257 1/587 35/0018-56/73-7/07 Table 8 0/0137 0/0129 f-tfgf tf i observation R² v 4/324 8/697 Table 9: Correlation Series Test-Pagan method f-tfgf tf i observation ² v R COPY RIGHT 2012 Institute of Interdisciplinary Business Research 504