Finansal varlık fiyatlama modelinin Türk sermaye piyasasında testi

Abstract

D2ET Finansal literatüründe son yıllarda üzerinde yoğun çalışmalar yapılan, finansal varlık fiyatlama modeli, risk ve getiri arasında ki ilişkiyi belirleyen bir modeldir. Bu modelin geçerliğini araştı ran birçok araştırmacılar modelin geçerliliği konusunda tam bir fi kir birliğine varmış değillerdir. Sermaye piyasaları çok gelişmiş ülkelerde yapılan testler, hem modelin geçerli olup olmadığı hem de sermaye piyasalarının etkinliklerini araştırmışlardır. Bu çalışmanın 2. bölümünde finansal varlık fiyatlama modelinin (FVFM) teorik geçmişi olan, portföy ve sermaye piyasası kuramları tartışılmıştır. 3. bölümde finansal varlık fiyatlama modelinin te melinde yatan varsayımlar, iki ana kısma ayrılarak incelenmiştir. 4. Bölümde ise FVFM'ni test eden çalışmalar ve elde edilen bulgular özetlenmiştir. 5. Bölümde ise finansal varlık fiyatlama modelinin Türk Sermaye Piyasası verileri ile test edilmesi yer almaktadır. Model test edilirken dört hipotez kullanılmıştır. Bunlar sırasıyla şöyledir : a) Finansal varlık verimi ile risk arasındaki ilişki doğrusal dır. b) Finansal varlık riskini en iyi beta temsil eder, bunun dışın da hiçbir risk faktörü bir finansal varlıktan beklenen verimi sistematik olarak etkileyemez. c) Modelin sabiti, tfo ile gerçekleşmiş değeri olan risksiz faiz aranı, R birbirlerine eşittir. d) Modelin eğimi, tf., gerçekleşmiş değeri olan Rm -Rf'ye eşittir. Bu çalışmanın bulgularına göre, H1 ve H2` hipotezleri kabul edi liyor. Bir başka ifade ile, risk-getiri ilişkisi doğrusaldır ve riski en iyi temsil eden ölçütü ise betadır. H3-, hipotezi ise 3. ana liz dönemi hariç diğer üç dönemde de kabul edilmiştir. Ancak, H4, hi potezi, 4. dönem hariç diğer üç dönemde de red edilmiştir. Bir başka ifade ile modelin tahmin ettiği Y1 değeri ile Rm-Rf arasındaki fark ol dukça anlamlı çıkmıştır. Ayrıca pazarın net getirişi olan Rm -Rf` dördüncü dönemin dışında sürekli negatif olmuştur. Bu çalışmanın bulgularına göre, test edilen modelin doğrusal olduğu ve finansal varlık getirilerini sistematik bir biçimde etki leyen faktörün yalnızca beta olduğu ispatlanmıştır. Ancak modelin bu sermaye piyasasında geçerliliğini test eden H3 ve H4 hipotezlerinden yanlızca H3 'ün kabul edilmesi, modelin et kinliğine gölge düşürmüştür. Ayrıca Rm - Rf 'nin genellikle negatif olması, Türk Sermaye Piyasasındaki yatırımcıların yüklendikleri daha yüksek risk karşısında ödüllendirilmedikleri ortaya çıkmıştır. - ıx -

SUMMARY THE TEST OF THE CAPITAL ASSET PRICING MODEL IN TURKISH STOCK MARKET The Capital Asset Pricing model, which is a well konwn model in finance literature, determines risk-return relation of capital assets. It is a version of a much older valuation technique-the risk premium model. The risk-premium model essigns increasingly high returns far increasing risks. An investment with the lowest possible return would be an investment with no risk. The return on this risk-free investment compansates the investor for his or her illiquidity (having the invested funds tied up) Dvei? the life of the investment. The capital asset pricing model is different from the typical risk-premium method because the CAPM measures risk in a particular way and must rest on some very explicit assumptions. These assumpt- sion have prompted abundant controversy. This study particularly concerns about testing the CAPM by using Turkish stock market's data, Before testing the model, we tried to explain the theoretical background of the CAPM. In section 2 we have facesud on portfolio valuation theories ranging from Markowitz's portfolito theory capital market theory and capital asset pricing model. It has been discussed Markowitz's meanvariance model and its particular problemns that practitioners had faced. And also, in this section, both portfolio and capital asset pricing models were discussed. The differences between capi tal market line were emphisized. Systematic and unsystematic risks and characteristic lire were also elaborately discussed. In Section 3 we have stated the assumptions that are lied behind the CAPM. hie have classified the CAPM' s assumption into two groups : a) Efficient market assumptions and b) Capital asset pricing model assumptions. The former group of assumptions indued, portfolio theory's assump tions as well. Eight assumptions were accepted. - x -Efficient-market's assumptions are : 1- The investor's objective is to maximise the utility of wealth. 2- Investors make choices on the basis of risk and return. Return is measured by the mean returns expected from a portfolio of assets;. risk is measured by the variance of these portfolio returns. 3- Investors have homogeneous expectations of risk and return. k- Investors have identical time horizons. 5- Information is freely available to investors. Capital asset pricing model's assumption : 6- There is a risk-free asset, and investors can borrow and lend at the risk-free rate. 7- There 'are no taxes, transaction costs, or other market imperfections. : 8- Total asset quantity is fixed, and all assets are marketable and divisable. In this section, uie have also discussed the relaxing of assump tions, and gathered some studies that were related to relaxing assumptions. Even more controversial than the model's basic assumptions are the results of sophisticated studies that have been developed to test the logic and realism of the CAPM. In section k me have discussed these various studies to determine arguments and findings most appropriately support or call into question the capital pricing model. üJe have grouped these studies into two main classes. In the first class, vthere are such studies that are related to misspecification of the model. As the model's form is simple, it is vulnerable to two potential sources of error. The first potential problem is that the form of the model may simply be wrong. Instead of being linear, the actual risk-return relationship could be non linear (for example, the `true` market line could be 3? staped). Such models might be misspecified. The second potential 'problem is austerity; the model may not include all the relevant factors. If a certain factor (or factors) does inflience the way that investors determine the price of an asset, and if this factor (or factors) is not included in the simple CAPM, the model would be termed inadequate to describe the real behavior of investors. - xi -In section 5 there is our emprical study which is testing the CAPM with Turkish stock market's data. The purpose of this study was to test emprically the risk return relationships for a mean- variance capital asset pricing cmgdel. To date all emprical work has focused on the Sharpe-linter [5>71 mean-variance CAPM. In this model, the risk of an efficient portfolio is measured by the standart deviation of return, &. For individual securities, the appropriate measure of risk is ?he covariance of return on the security and the marka t portfolio. The CAPM states that the expected return of any security or portfolio equals the risk-free rate of return, R_, plus a risk premium that is (beta) times the difference between the expected return on the market portfolio and the risk-free rate of return, i.e., E(R.) = Rf + [E(Rm) - Rf] /3. Evidence presented by Linter [s], Miller and Scholes [5l], Black, Jensen and Scholes [52] Fama' and Macbeth /k9/ seems. td indicate that' the model does' not provide a complete `decription. of security returns» In this study the model was tested with regard to following four hypotheses : (H ) The relationship between the expected return of a security and its risk in. an efficient portfolio is linear. (H ) Beta (;â).:i-s a complete measure of risk of a security in the efficiend portfolio; no other risk factor systematically affects the expected return on a security. (H3) The value of the intercept, tf, is equal to the risk- free rate of return. ° (H^) The slope of the Model, if, is equal to the E(R )- R, and higher risk should be1associated with higher return. To test these four hypotheses we have used kk firms commons^ tocks which are registered in Istanbul Stock Exchange. üJe have used montly data for the period from I960 to 1987. The CAPM test procedure was similar to the one used by Fama and MacBeth [k3]. The analysis consisted of two regressions. In the first regression, ft and the variance of error, S(e.), terms were estimated by regressing the realized rate of return '`'on security against the realized rate of return on the market portfolio, i.e., xn -R..= e*. + A, R,+ E., it i *Ji mt it t = 1,..., 36 i = 1,.... ki* The standart deviation of the error term, S(e, ), is a measure of risk that is not related directly to (3,. In the second regres sion equation the realized rate of return in the next period was used as the dependent variable and was regressed against three explanatory variables, i.e., Rit= ^o + *i A,t-1+ ^2,/3i,t-l + V^i.t-l* + Bit üJehere t, y., Y` and X-, are least square estimates of the regression coefficients, n.is the estimated error term, andj3-., and S(e.,. -i ) are obtained from first regression equation. Tne` variable' ~J3? +., was included to test for linearity. H, is not rejected if X?=D. The variable S(e.., ) was included to test (H`). Hypotesis (H`) is not rejectedif yy=Q. The test second cross-sectional regression equation, we must eliminate or, at least, decrease the measurement errar bias. That is why we need to use `true` values of the relative risk measure â', but in emprical test estimates of fii must be used. But these are subject to measurement errors. If the measurement errors in $X. are less than perfectly positively correlated,^ of a portfolio may be a more precise estimate of the true ]3/ than ji is of the truejâ. for individual securities. ^ In forming portfolios, the securities should be grouped to obtain the maximum possible dispersion of the risk coefficients (ftD- The mare dispersed the values of Â, the smaller variance of^the estimated coefficients. ^ Forming portfolios on the basis of the ranked values of /§. for securities, could introduce a selection bias. The selection bias can be overcome to a large extend by forming portfolios from ranked /§. computed from data far one time period and the using a subsequent period to obtain i§. In this study, the approach used to compute was as fallows : Data from the first period (1980-82) were utilized to compute^.. The /3 s were ranked in ascending order and were used to form II (in forth period 9) portfolios each consisting of h common stacks. The first k common stacks formed the first portfolio. The second k cammonstacks formed the second portfolio, ect. Data from the subsequent period (1981-B3) were usedJ;o estimate /3, ft* and S(e ). R * was the average of the squared fi. values in' ^ ^ - xm -portfolio p;and S (i ) was average S(e.) in portfolio p. The Expec ted returns in the next period (1962-84) are regressed against the explanatory variables^, ^3xand S(e ) estimated in the 1981-83 period. P P. P The major results of the model are presented in Table 5.5. To give a more detailed presentation of the structure of risk and return, the results are reported for four versions at the risk- return regression equation. This equation is, RPt= V *ıt/p,t-ı+ W,t-ı+ ^S,^ + BP,t First panel centains only one explanatory variable, >S '. second panel contains two variables,^ t -, and^â2., ^tnird panel contains two variables^,, and'aCe,, ) ; fourth panel contains all there variables %L 7 -,, ğz, P'ana S(e,,). /°p,t-l' r p,t-l Pit-1 A For each model and period, the table lists %., the regression coefficient estimates t (y.), the t-statistics testing H : £.= D, and R-square and F-statist5cs. ^ The finding of this study can be summarized as follows : As reportet in fourth panel, Hypates is H. (H £&,,=Q), which states that the relationship between expected return and jâ is linear, was supported in all analysis period aW= D.G5. ^Likewise, as indicated in third and fourth panels, hypotesis H`(H :^=D) was not rejected. Therefore the evidence indicates that no measure of risk other than systematically affects expected returns. These results consistent with Black, Jensen and Scholes [52] and Fama and MacBeth [kS]. Since/3Z,, and S(e,, ) were found to be nonsignificant, we have deleted them frorrrcro'ss-sectional regression equation. Thus, the potential multicallinearity problem among the explana tory variables could be reduced. After eliminating two variable from regression, the new:reg- ression we have used is, fipt=^at+^lt^p,t-l+'V t= 1,..., h This equation was used to test (H ) and (H,). If traditional CAPM is walid, ?Dt= Rft and ^fc= R^- tfft# The results of thiâ.'study with respect to hypotesis (H ) and (H, ) were summarized in Table 5.6. - xxv -The estimated intercept of model, tf was less than its rea lized value ^except in fourth analysis period. But the differen ces between tf. and R`, mere not significant at °(- 0.05 level. To test H (H ? Rff.= $ 4.) we have computed t(R ) statistics. In Table 5.6, out of thira period, in three periods t (Rf) values were belove t(c<=D.D5). It means hypotesis H, was accepted. Namely, risk-free interest rate equals interecept which was esti mated by CAPM. As to hypotesis H,, the results were not as clear as H,. To test H, we have caculated t(R ) statistics in table 5.6. this statitcs was greater than the value of t (c<=0.05) statistic except forth analysis^ period. It means the H, is rejected and the differences between £, and its realized value, R.- R`, could,,. `, It mt ft not be ignored. The finding of this study relating to hypotesis H-. and H, states that the CAPM may not provide an appropriate description of the relationship between risk and return. The source of the problem may be the assumption of unlimited riskless borrowing opportunities or errors of measurement Df Rf. Another importent results were that in Turkish stock market, investors are not compensated for taking a higher risk. Because the differences between R and R` are negative in all periods, except fourth period. - ?- The results of this study may suggest that the CAPM does not appropriate to interpret the relationship between risk and return in Turkish stock market. And another important result is that Turkish stock market is not efficient as far as the inves tors are concerned. Before coming to these conclusion we must keep in mind that the results of this study have affected by scarcity of.: data about Turkish stock market. We have used only kh firms commons- tacks for eight years. In comparing to other studies which were made in developed countries, our data is not sufficient enough. But this problem is a common problem in less-developed or develo ping countries in which the capital markets are thin [65, 66, 67, 68, 69]. But the development in Istanbul Stock Exchange is promising that in a few years the number of commenstocks, in the market will be increasing. - xv -Besides the scarcity of data, relability of data is also very important-:. To overcome this problem, the authorities should arrange new departments in stock exchange that provide periodic data about firma activities. UJe have tried to do; a pioneering study questioning some issues which are related to the CAPM and Turkish Stock Market. Ide would be very happy if we could contribute something to this field. - xvx ~