An Empirical Study on the Closed-End Funds Performance in China
by user
Comments
Transcript
An Empirical Study on the Closed-End Funds Performance in China
An Empirical Study on the Closed-End Funds Performance in China using Conditional Measures Yang Kuan College of Business Management, HuNan University, Changsha, P.R.China,410082 Abstract Most studies of fund performance use unconditional measures that are susceptible to bias caused by common time variation in risks and risk premia. We evaluate the closed-end fund performance of China using information variables that has produced new insight about empirical analysis of the fund performance in China. The results indicate the trading volume from China stock exchange is better conditioning information, the performance evaluation using information variables: trading volume, Jensen alpha as are higher when estimated with the conditional model and the number of significant timing coefficients is greatly reduced. Keywords Mutual funds; Performance evaluation; Conditional CAPM 1. Introduction Fund manager performance is typically measured relative to some ‘normal’ or ‘expected’ performance. The predominant definition of ‘normal’ applied in performance studies is based on the traditional CAPM which assumes that fund risk and expected returns are stable over time. If fund risks and risk premia change over time, however, the traditional performance measures will confound time variation with abnormal performance. Ferson and Schadt(1996)advocate using performance measures that are conditioned on public information variables in order to avoid the bias induced by using historical average returns to estimate expected performance[1]. Their results demonstrate that using conditioning information is both statistically and economically significant. The average poor performance and perverse market timing ability that is commonly indicated by unconditional performance measures is negated using the conditional model. This paper examines the effect of trading volume as lagged information variables in the analysis of investment performance using weekly data for China closed-end funds over the period 2000–2004. The study provides corroborating results to those of Ferson and Schadtand, Ferson and Warther(1996) [2]. Estimating conditional Jensen alphas causes the distribution of alphas to shift to the right towards positive values, making the sample of funds look better. The traditional Treynor and Mazuy(1966)measure of market timing[3] produces evidence that an average fund takes on more market exposure when market returns are low. We find here that the conditional market timing model reduces the number of significant negative timing coefficients. 2. The conditional models The standard performance measures that are applied extensively in evaluating managed fund performance were proposed more than three decades ago. The traditional CAPM was widely used as the benchmark model to measure risk-adjusted portfolio performance (e.g., see Jensen,1968)[4]. The risk-adjusted performance of managed portfolios, α p , then was given by: rp ,t +1 = α p + β p rm ,t +1 + ε p ,t +1 , t = 0,K , T − 1, p = 1,K , N where rp ,t +1 is the excess return on portfolio p between t and t + 1 , β p is (1) the sensitivity of the excess return on the fund to the the excess return on the market portfolio, rm ,t +1 is the excess return on the 618 benchmark portfolio between t and t + 1 , ε p ,t +1 is the random error of fund p in weekly t + 1 . Manager ability is measured by what has become known as the ‘Jensen alpha’. Superior or inferior managers would have consistently positive or negative. In specifically considering managers’ ability to predict market moves, Treynorand Mazuy(1966)add a quadratic term in another model based on CAPM that has become a standard for measuring timing ability. rp ,t +1 = α p + β p rm ,t +1 + γ p rm2,t +1 + u p ,t +1 , t = 0,K , T − 1, p = 1,K , N where (2) α p is a measure of timing-adjusted selectivity, β p is the beta, γ p is the market timing coefficient. Positive alpha and gamma values indicate that the Manager has superior selection and timing skills, respectively. Multi-index models employed for example by Elton et al.(1993)improve the accuracy performance measurement by controlling for style choice[5]; however, both single- and multi-index models suffer from another problem: time-variation in risks and expected returns that may be misinterpreted as superior selectivity or timing skills. Performance measures that use unconditional expected returns calculate abnormal performance as the difference between the average portfolio excess return and a beta-adjusted average market index excess returns. If the market risk premium changes and the performance metric does not control for this, time variation in the market risk premium will be reflected in the estimate of abnormal performance and mistaken for manager under- or over-performance. Ferson and Schadt(1996)argue that evidence of return predictability using predetermined variables represents changing required returns. They propose a modification to the Jensen alpha and market timing models to incorporate conditioning information that allows for the estimation of time-varying conditional betas. When incorporating lagged information variables on dividend yields and interest rates, they find that the conditional model improves measured performance. Unconditional alphas indicate average poor performance, however the distribution of conditional alphas is consistent with neutral performance. They also find that evidence of perverse timing is removed when using the conditional market-timing model. Ferson and Schadt(1996)modify the traditional Jensen alpha model(Eq.(1))by adding a vector of lagged public information variables to equation to estimate α p the conditional measure of performance: u p ,t +1 = α p + δ1 p um ,t +1 + δ 2' p ( zt um ,t +1 ) + ε p ,t +1 (3) where u p ,t +1 and um ,t +1 are the excess return on portfolio p and on the market index over the risk-free rate δ1 p is the parameter estimating the conditional beta of portfolio p; zt is the vector of public information variables at time t; δ 2 p is the vector of parameters that measure how the conditional beta varies at time t+1; with respect to the vector of market indicators. The conditional measure of timing, γ tmc , is estimated modifying Eq.(2): u pt +1 = a p + bp ρ umt +1 + C p ( zt umt +1 ) + γ tmc [um ,t +1 ]2 + v pt +1 (4) The public information variables, zt , represent information available at time t for predicting returns at time t+1. These interaction terms pick up movements of conditional betas as they relate to market indicators. The coefficients, C p , estimate the response of conditional betas to lagged market indicators. By capturing information available to managers at t, the vector [ zt × umt +1 ] precludes strategies that can be replicated using public information from being ascribed with superior selectivity or timing ability on the basis of this information. The statistical reason for this is that the interaction terms measure the covariance between 619 conditional beta and the expected value of the market return using lagged instruments. The difference between unconditional and conditional alpha is determined by the average values of the interaction terms. Positive(negative)covariance will result in lower(higher) abnormal performance. The purpose of this study is to examine the performance of closed-end funds in China using conditional models. The tests performed here provide insight into the effect of using conditioning information in measuring the performance of China closed-end funds. Anther purpose is to indicate the trading volume from China stock exchange is better conditioning information variable. 3. Data and Sample In China, only the first 20 listed funds can provide above five year data. We take the period of fund data existed from January 1,2000 to December 30, 2004.The observational period covering the latest five years and better reflect the reality and the regulation of the securities market and the fund industry in China. In observational period, the securities market in China becomes much stronger and rational with the increasingly rapid growth. As for the sample, we include all longer 20 closed-end equity funds. Now listed funds are all closed-end funds. According to the net asset per share publicly published every week, the rate of the log returns can be examined as NAVit log Rit = log NAVit −1 (5) , where log Rit is the log return on fund i during week t NAVit is the net asset of the i fund per share during week t. Confined to the governing regulation, the stock holding of the fund portfolio should be no more than 80 percent. And the bond and cash section should be no less than 20 percent. So in addition to directly use the marked mentioned above, we construct a new benchmark is composed of 40 percent of returns on the Shenzhen A-stock Component Index and the Shanghai A-stock Composite Index separately, and 20 percent of the cash and bonds yield (it is called the mix market index here after). One-year deposit interest rate is used as risk free rate. However, the central bank decreased the interest rate for the two times on 99.6 and 02.2 separately. Accordingly the weighted adjusted weekly risk free rate we employ in this is 0.00039 (interest tax-free). And we replace the cash and bonds yield with the adjusted weekly risk free rate. The rationale for the equal weights of two market index in the new mix index is the highly correlation between these two market, the correlation coefficient of the returns on Shenzhen and Shanghai stock market in our sample is 0.9326. The data of the index consists of the closing price on each Friday during sample period. As market indicators that previous studies identify as predictors of security risks and returns over time. The variables employs (1) the lagged level of the one-month Treasury bill yield, (2) the lagged dividend yield of the CRSP value-weighted New York Stock Exchange (NYSE)and American Stock Exchange(AMEX) stock index, (3) a lagged measure of the slope of the term structure (4) a lagged quality spread in the corporate bond market and (5) a dummy variable for the month of January. As we is in developing country, these variables don’t fit to predict of security risks and returns over time. We use the trading volume yield from China stock exchange at the end of the previous week files to predict the future returns. This is based on ‘it takes volume to make prices move’. , , 620 rit +1 = α + βiTOTOt + ui ,t +1 (6) Where TOt variable represents turnover. According to the Eq.(6), we can construct the regression, as shown in table 1. Table 1 shows that the turnover variable and return on funds have significant relation, so we believe the trading volume yield from China stock exchange can predict the future returns. Table1 Regression estimate parameter ad R β 2 0.16 T -Statistic 0.018 6.1 P -value ** 0.00 *Significant at the 1% level. **Significant at the 5% level. 4. Empirical Results and analysis Data are used in an ordinary least-squares(OLS)regression to estimate the unconditional Jensen model Eq.(1)and the conditional model Eq.(3). The conditional model results are reported in Table 2. There is evidence that the conditioning variables are individually and jointly significant, supporting the proposition that the trading volume variables are related to excess fund return. In the all-funds sample tests, abnormal return of 18 funds are significant at the 1% level, while fund JINGHONG are not significant. The results provide evidence of the marginal explanatory power of conditioning information in the performance measure. The adjusted R-squares are slightly higher for the conditional model. These results are similar to those of Ferson and Schadt(1996). Table2 Fund α JINTAI 0.0018 TAIHE 0.0026 ANXIN 0.0021 HANSHENG 0.0018 YUYANG 0.0019 Conditional Jensen measure of performance of 20 Samples t -Statistic 3.44 ANSHUN 0.0027 3.55 XINGHE 0.0027 KAIYUAN 0.0021 PUHUI 0.0019 TONGYI 0.0025 JINGHONG 0.001 YULONG 0.0021 ** 2.16 0.0025 0.0023 5 ** 0.0022 0.0028 0.48 ** 2.47 XINGHUA JINXIN 17 ** 4.26 JINGYANG YUYUAN δ1 ** 2.44 2.5 Rank 2.11 0.41 0.47 15 0.44 17 9.3 13 9.2 10 0.56 ** 7 0.37 12 ** 4 0.5 12 2.83 3.89 1.86 ** 8 0.6 ** 2 0.5 ** 3 0.56 * 13 0.49 ** 16 0.44 ** 3.8 1.35 2.37 11 18 12 ** 4.12 2.94 0.52 t -Statistic ** 6 0.44 20 0.49 12 0.51 621 7.9 15 8.6 14 5.1 10 14 13 13 ** ** δ2 -0.04 -0.06 ** -0.06 ** -0.06 ** -0.05 t -Statistic -2.3 -3.4 -4 ** 0.71 ** 0.78 ** -3.2 -2.8 0.66 ** 0.61 -0.05 ** -0.03 -1.7 -1.1 ** -0.08 -4.4 -0.03 ** -0.04 ** ** ** -0.07 * -2 0.59 ** ** ** R2 ** 0.67 0.46 0.7 ** 0.84 * 0.62 -0.04 -1.7 -1.4 0.76 -0.04 -1.5 0.58 ** -0.05 ** -0.03 ** -0.04 -3.6 ** 0.6 ** 0.68 -3.2 -1.4 -1.7 * 0.67 0.66 PUFENG 0.002 JINGBO 0.0017 TIANYUAN 0.0029 TONGSHENG 0.0023 3.23 2.62 3.1 ** 14 ** ** 4.02 ** 0.53 19 0.5 1 0.55 9 0.52 15 17 9.7 21 ** -0.04 -1.8 ** -0.04 ** -0.06 -3.3 ** -0.05 -3.7 -2.1 * 0.73 ** 0.7 ** 0.71 ** 0.75 *Significant at the 1% level. **Significant at the 5% level. Table 3 provides evidence on whether the conditional alphas are significantly different from the traditional alphas. The number of alphas produced by the traditional measure is all smaller than condition’s alphas. On the other hand, the number of t-statistic produced by the traditional measure only 4 of 20 superior to the conditional model t-statistic. This implies that when public information is incorporated into the performance measure, the distribution of the alphas shifts to the right and consequently makes the sample of funds look better. The results are consistent with Ferson and Schadt(1996)and Ferson and Warther(1996), that the conditional alphas are higher than the traditional alphas. Table3 Fund Conditional approach Traditional approach Conditional t -Statistic Traditional t -Statistic Fund Conditional approach Traditional approach Conditional t -Statistic Traditional t -Statistic JINTAI Traditional and conditional abnormal performancesof 20 Sample HAN YU JING XING AN TAIHE ANXIN SHENG YANG YANG HUA SHUN JINXIN YU YUAN 0.0018 0.0026 0.0021 0.0018 0.0019 0.0022 0.0025 0.0027 0.0023 0.0028 0.0015 0.0022 0.0017 0.0013 0.0015 0.0019 0.0023 0.0021 0.002 0.0025 2.44 4.26 2.47 2.5 2.16 2.11 3.44 3.55 4.12 2.83 2.468 3.85 2.454 1.848 2.086 2.347 2.791 3.112 3.756 3.174 XING HE KAI YUAN PUHUI TONG YI JING HONG YU LONG PU FENG JING BO TIAN YUAN TONG SHENG 0.0027 0.0021 0.0019 0.0025 0.001 0.0021 0.002 0.0017 0.0029 0.0023 0.0024 0.0018 0.0013 0.0021 0.0008 0.0018 0.0017 0.0014 0.0024 0.0019 3.89 1.86 2.94 3.8 1.35 2.37 3.23 2.62 3.1 4.02 3.812 2.14 1.839 3.375 1.177 2.409 2.563 2.063 3.406 3.066 Timing coefficients in the traditional and conditional Treynor–Mazuy models Eqs.(2)and (4)are estimated using individual funds. Ten of the 20 individual funds’ timing coefficient estimates are negative, don’t indicating perverse timing where the manager lowers exposure to the market when the market performs well and increases exposure in poor markets. Of these 10 estimates, only 3 are significant at the 5% level. Under the conditional version, 5 of the 20 estimates of the conditional timing coefficients of the individual funds are negative. Of these 5 estimates, none are significant at the 5% level. Consistent with Ferson and Schadt(1996)and Ferson and Warther(1996), the evidence of perverse timing ability is removed under the conditional model. 622 Table4 Unconditional Treynor-Mazuy measure of performance of 20 Samples Fund α JINTAI 0.0016 2.23 TAIHE 0.0024 3.82 ANXIN 0.0017 t -Statistic 2.15 ** ** ** ** HANSHENG 0.0018 2.38 YUYANG 0.0014 1.72 * JINGYANG XINGHUA ANSHUN 0.0013 0.0018 0.0024 1.42 2.01 3.18 JINXIN 0.0019 3.16 YUYUAN 0.0019 2.12 XINGHE 0.0023 KAIYUAN 0.0005 PUHUI 0.0022 TONGYI JINGHONG YULONG 0.0025 0.0008 0.0018 3.13 2.76 3.54 2.23 3.29 JINGBO 0.0019 2.53 TONGSHENG 0.0024 *Significant at the 1% level. **Significant at the 5% level. ** ** ** ** ** 0.99 0.0024 0.0021 ** 0.52 PUFENG TIANYUAN ** 2.65 3.52 ** ** ** ** ** β t -Statistic 0.4648 21.1 0.5054 25.4 0.383 15.8 0.4523 18.6 0.4208 16.5 0.533 0.3599 0.476 19 12.6 19.9 30.6 0.4809 17.4 23.8 0.4598 15.9 0.425 16.8 0.4269 0.4776 0.4939 19.6 19.5 19.1 0.518 22.7 0.4896 21 0.5213 0.51 ** ** ** ** ** 0.5878 0.5388 ** ** ** ** ** ** ** ** ** ** ** ** ** 20.6 23.8 ** ** γ t -Statistic R2 -0.034 -0.08 0.7 -0.304 -0.83 0.77 0.0203 0.045 0.57 -0.697 -1.56 0.64 0.1439 0.307 0.59 0.7698 1.49 0.66 0.4971 0.942 0.46 -0.405 -0.92 0.67 0.1402 0.397 0.83 0.7768 1.53 0.63 0.1984 0.476 0.75 ** 1.6208 3.034 -1.088 -2.33 -0.482 -1.2 0.67 0.0185 0.041 0.67 -0.081 -0.17 0.66 ** ** 0.59 0.59 -0.878 -2.09 0.73 -0.647 -1.5 0.7 0.3758 0.807 0.7 -0.661 -1.68 0.75 From table4 and 5, that provides evidence to examining whether the conditional timing coefficients are significantly different from the traditional timing coefficients. The t-test indicates that for the entire sample, conditional timing coefficients are significantly different from the traditional timing coefficients. It should be noted that the number of significant negative timing coefficients are greatly reduced when the conditional model is used. This is consistent with Ferson and Schadt(1996)and Ferson and Warther(1996). 623 Table5 Conditional Treynor-Mazuy measure of performance of 20 Samples Fund JINTAI α 0.0015 TAIHE 0.0024 ANXIN 0.0016 HANSHENG 0.0018 t -Statistic 2.6 3.8 2.1 2.3 ** b 0.4797 ** 0.5247 ** 0.4077 ** 0.4711 YUYANG 0.0013 1.7 * 0.4425 JINGYANG 0.0012 1.4 0.5582 XINGHUA 0.0018 ** 0.3763 2 ANSHUN 0.0023 JINXIN 0.0019 3.1 YUYUAN 0.0018 2.1 3.2 18 0.0022 2.7 0.5572 21 24 0.4899 17 ** 0.4418 17 ** 0.4431 TONGYI 0.0024 JINGHONG 0.0007 3.5 0.9 YULONG 0.0018 2.2 PUFENG 0.0024 JINGBO 0.0019 3.5 13 30 PUHUI 2.6 19 0.5021 3.1 0.4 0.0024 17 ** 0.0022 0.0021 19 0.6019 0.0004 TIANYUAN 17 0.5033 XINGHE TONGSHENG 26 ** KAIYUAN 2.5 12 ** ** 3.3 t -Statistic 20 0.4894 19 ** 0.5095 19 ** 0.5242 22 ** 0.5002 21 ** 0.5479 ** 0.5243 21 24 ** C -0.05 ** -0.065 ** -0.083 ** -0.063 t -Statistic -2.5 -3.3 -3.5 -2.6 ** γ t -Statistic R 2 0.486 0.63 0.71 ** 0.372 0.91 0.79 ** 0.8809 1.76 * 0.6 ** -0.042 -0.1 0.66 ** 0.9002 1.7 * ** -0.073 ** -0.085 -3.1 ** 1.6503 2.83 ** 0.68 ** -0.055 -1.9 * 1.0697 1.77 * 0.47 ** -0.091 ** 0.5472 1.12 0.7 ** -0.047 -2.5 ** 0.6322 1.57 0.84 ** -0.071 -2.6 ** 1.5156 2.63 ** 0.8399 1.78 * 0.76 ** 2.672 ** 0.62 ** -0.501 ** -2.5 -1.6 0.0819 0.18 0.68 0.428 0.83 0.67 ** 0.4614 0.85 0.67 -0.659 -1.4 0.73 0.7 -2.9 -3.9 ** -0.062 ** -0.101 -3.6 ** -0.056 -2.2 ** -0.054 ** -0.039 ** -0.052 ** -0.021 -2 -0.9 ** -0.036 -1.5 ** -0.089 ** -0.048 -2.7 -3.6 -2.2 ** 4.48 -0.9 -0.277 -0.6 ** 1.3021 ** ** -0.163 2.5 -0.4 0.61 0.64 0.6 0.71 0.76 *Significant at the 1% level. **Significant at the 5% level. 5. Conclusion From the above empirical investigation, this paper researches the following conclusion. This study examines the effect of trading volume information variables in the evaluation of investment performance using China closed-end funds data, using of conditioning information in performance is statistically significant. Furthermore, using conditioning information improves the performance of the funds, causing the distribution of the alphas to shift to the right. We find that the number of significant negative timing coefficients is greatly reduced after incorporating public information variables into the model. Finally, the tests in this study employ conditional versions of only the single-index model. An important application of the Ferson–Schadt framework is to alternative models of equilibrium such as that used by Elton et al. (1993)which includes the excess returns on an equity index of non-S&P small stocks and on a bond market index to represent the performance of assets typically held by fund managers but not represented in the S&P index. 624 References [1] Ferson, W.E., Schadt, R., Measuring fund strategy and performance in changing economic conditions. Journal of Finance, 1996,51:425–462. [2] Ferson, W.E., Warther, V.A., Evaluating fund performance in a dynamic market. Financial Analysts Journal, 1996,52(6): 20–28. [3] Treynor, J., Mazuy, K., Can mutual funds outguess the market? Harvard Business Review, 1966, 44:131–136. [4] Jensen, M.C., The performance of mutual funds in the period 1945–1964. Journal of Finance, 1968,23:389–416. [5] Elton, E.J., Gruber, M.J., Das, S., Hlvaka, M., Efficiency with costly information: a reinterpretation of evidence from managed portfolios. Review of Financial Studies,1993, 6:1–22. 625