بتا مشروط: شواهد حاصل از بازارهای نوظهور آسیا
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|14065||2011||24 صفحه PDF||سفارش دهید||16413 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Global Finance Journal, Volume 22, Issue 2, 2011, Pages 130–153
Pettengill, Sundaram, and Mathur (1995) respond to the prima facie failure of the standard CAPM and propose a conditional beta model by segmenting the market into two states – up markets (where the market excess return rm–rf is positive) and down markets (where rm–rf is negative). We examine this model in eleven Pacific Basin emerging markets using a range of variants: a model where betas are calculated using local excess returns, a model where betas are calculated using world excess returns, a model using both local and world excess returns and a model using both local and world excess returns where local returns are orthogonal to world returns. Only in the last of these formulations is there some evidence supporting the conditional beta model.
A central tenet of standard finance theory is the positive relationship between risk and expected return. The Capital Asset Pricing Model (CAPM) (Black, 1972, Lintner, 1965 and Sharpe, 1964) argues that systematic risk, measured by beta, is the only relevant risk measure. Early empirical evidence supported the validity of CAPM (see, for example, Black et al., 1972 and Fama and MacBeth, 1973). However, following the seminal work of Fama and French (1992), there has been mounting evidence suggesting that beta is an incomplete measure of risk or even that it bears no relationship to returns (see Fama & French, 2004, for useful background). The CAPM is a theoretically elegant and tractable model that, almost twenty years after the publication of Fama and French (1992), is still used in empirical research and industry applications. Proponents of the CAPM have been reluctant to accept its death and a number of theoretical and empirical counterblasts have been proposed.1 One response to the prima facie empirical failure of the CAPM is provided by Pettengill et al. (1995) (hereafter PSM). PSM propose a modified CAPM to explicitly recognize the impact of using realized returns as a proxy for expected returns. PSM argue for a conditional relationship between beta and realized returns: a positive relationship between beta and realized returns in up markets (where the realized market return exceeds the risk-free rate of interest), and a negative relationship between beta and realized returns in down markets (where the realized market excess return is negative). PSM test their hypothesis and find a significant and systematic relationship between conditional beta and returns for the US market. PSM's approach has recently captured the attention of a number of academic studies. Fletcher (1997) examines the UK stock market for the period 1975–1994 and finds a significant relationship between beta and returns, although the estimated up-market and down-market risk premiums differ in magnitude. Fraser et al., 2004 and Hung et al., 2004 also examine the UK stock market; Isakov (1999) the Swiss stock market; Elsas, El-Shaer, and Theissen (2003) the German stock market; Theriou et al., 2004 and Theriou et al., 2007 the Greek market; Karacabey and Karatepe (2004) the Turkish market; and Sandoval and Saens (2004) the markets of Argentina, Brazil, Chile and Mexico. All these studies report a significant relationship between beta and returns using the PSM approach. Instead of using portfolio returns, Fletcher, 2000, Tang and Shum, 2003a and Tang and Shum, 2003b employ aggregate indices data on international markets and also find evidence supportive of the PSM hypothesis.2 Evidence against the conditional beta model may be found in Lilti and Montagner's (1998) examination of the French market. There are five papers that test the PSM hypothesis using either a single or a pair of Asian emerging countries. They all find evidence suggesting that the conditional beta model is useful in explaining returns. Faff (2001) examines the Australian market using portfolios based on industries and finds some, but not full, support for PSM: it is primarily industrial stocks, in contrast to the resources sector, which drive his findings. Early research for Hong Kong finds support for PSM using an analysis based on ten portfolios formed by ranking on beta (Lam, 2001). Tang and Shum analyse Singapore (Tang & Shum, 2004), Hong Kong3 (Tang & Shum, 2006) and Taiwan and Korea (Tang & Shum, 2007) using portfolios based on beta (and in their paper on Taiwan and Korea, they also sort on size, as we do4). Given the pioneering nature of Tang and Shum's papers, they can only examine a relatively short time period (1992 to 1998) compared to the periods studied in this paper.5 Generally, their results are supportive of PSM's model but they also add other variables to the model and find these variables to be statistically significant. That additional variables that are significant should be of concern to proponents of the CAPM; like a jealous lover, the CAPM does not allow any “rivals”.6 We analyse PSM's model using data from eleven emerging markets from Asia using ten size-based portfolios (updated annually) for each market. Analyzing a large number of markets will perhaps allow us to consider not only whether the model is applicable in these markets but also to consider whether examination of the model may be warranted in other markets. Given that emerging markets are believed to experience higher levels of risk, we also believe that such markets provide excellent experimental settings to test a model that has the return to risk trade-off at its core. Size-based portfolios may also present a greater challenge to PSM's model if size represents a persistent anomaly within the CAPM framework (as Banz, 1981, would suggest). Visual inspection of the data, included in the appendix to this paper, suggests that size-based portfolios may pose a considerable challenge to the CAPM or PSM's variation of it. In the appendix, we plot the excess returns, Rp–Rf , and the market risk-premium, Rm–Rf, for the portfolio of the largest stocks and the portfolio consisting of the smallest stocks in each of the markets we study (as well as a simple linear regression for each portfolio). In all cases, the excess return of the portfolio of the largest stocks has a positive relationship with the market risk-premium and, as should be expected, the coefficients we report for beta (β) are all around one. In contrast, in each market, there appears to be a negligible relationship between the excess returns of the portfolio of the smallest stocks and the market. If there were a small firm premium, following Banz (1981) we might expect to see values of beta greater than one. However, the phenomenon of small firms behaving differently to large firms has been documented previously; in Australia, for example, Durand, Limkriangkrai, and Smith (2006) find that their model does not adequately describe the returns of firms with low market values. In addition to considering the pricing of beta calculated using local market excess returns, we also investigate a number of variations to the model. The first is where beta (both for the unconditional and conditional models) is calculated using a world market excess return (see Bruner, Li, Kritzman, Myrgren, & Page, 2008). Bruner et al. (2008) argue that the choice of either domestic market index or global index depends on the level of global market integration. More generally, global systematic risk may play a role in the returns of stocks listed in particular national markets ( Harvey & Bekaert, 1995). Consideration of the potential role of a global, rather than a local, beta in the cross-section of returns recognises the possibility of market integration. The second of these additional models considers whether both local and world market risk are priced using a model containing two betas: one calculated using local market excess returns and the second calculated using world market excess return (we call this the hybrid model). Bodnar, Dumas, and Marston (2003) propose a hybrid model arguing that both home market segmentation and world market integration prevail to a certain extent. The final model recognizes that, if markets are integrated, local markets may be driven by the world market. Therefore, we adapt the hybrid model and calculate betas using local market excess returns which are orthogonal to the world market excess return; we call this the orthogonalized hybrid model. Our empirical analysis, presented in Section 3 of the paper, finds evidence consistent with PSM's model using the orthogonalized hybrid model. However, when we consider if conditional betas are themselves conditional on state-dependent expectations (following Lewellen & Nagel, 2006) we find the model wanting. We also provide evidence that global systematic risk is priced in some markets (consistent with Bruner et al., 2008). Our study therefore provides support that world systematic risk must be considered in models capturing the cross-section of returns. Our analysis also highlights the importance of considering the potential asymmetry between up and down markets in asset pricing tests. Further, we provide evidence that asset pricing tests need to consider the information sets used by investors to condition their expectations. The remainder of the paper is organized as follows. Section 2 presents the data and methodology. We give details of our modified version of the two-pass procedures originally proposed by Black et al. (1972) and its application to three alternative versions of the conditional beta models. Specifically we use the Generalized Method of Moments (GMM) technique to deal with the errors-in-variables problem. Section 4 concludes the paper.
نتیجه گیری انگلیسی
Pettengill et al. (1995) (PSM) note the failure of the standard CAPM and propose a modified methodology that deals with the problem of using realized returns to proxy for expected returns. The data are segmented into up-markets, defined as periods where the market excess return (rm–rf) is positive; and down markets, defined as periods where the market excess return (rm–rf) is negative. This methodology has found some success in uncovering a systematic conditional beta-returns relationship in many markets around the world. This study applied the PSM (1995) segmentation procedure to eleven Pacific Basin emerging markets. We study a number of variations of the segmentation model: a model where betas are calculated using local excess returns, a model where betas are calculated using world excess returns (expressed in local currency), a model using two betas (one calculated using local excess returns and the other calculated using world excess returns) and a model where betas are estimated using local market excess returns which are orthogonal to the world market excess return. We called this last model the orthogonalized hybrid model. Tests of the standard CAPM reported in this study present results that are largely in keeping with those of Fama and French (1992): the standard unconditional beta does appear to be dead in the eleven countries we study. However, in three markets we study – Hong Kong, Korea and Thailand (and also marginally for Sri Lanka) – we find evidence that global systematic risk is priced. We find some evidence supportive of the PSM (1995) dual-hypothesis – that segmentation of the sample will lead to a positive estimated market risk premium in up markets and a negative estimated market risk premium in down markets – in our analyses of the orthogonalized hybrid model (Panel D of Table 8). The analyses appear robust when a size variable is included (Table 9); that is, the size variable adds little to our analysis. We also considered if the conditional betas we study are themselves conditional (that is, whether they vary dynamically as investors' information sets change) following the methodology of Lewellen and Nagel (2006) (Table 10). Although the analyses of the orthogonalized hybrid model reported in Panel D of Table 8 encouraged us, our analysis of the model using the approach of Lewellen and Nagel (2006) leads us to conclude that PSM's model is not a good candidate to model the cross-section of equity returns in the markets we study. Finding that PSM's model is unsupported is at odds with studies we have cited in this literature. Literature supporting PSM is, however, at odds with what might be considered the major stream of thought in empirical asset pricing following Fama and French, 1992 and Fama and French, 1993. The tests in this paper have focussed on emerging markets where returns may be more volatile than other markets; there may be too much noise to get clear readings. Unlike a number of studies that have supported PSM's model, we have also utilized size-based portfolios in the tests reported in this paper. Size was perhaps the first hurdle the standard CAPM failed to jump (Banz, 1981) and, as we noted in the Introduction to this paper, visual inspection of the data (included in the appendix to the paper) suggested that the behaviour we observed for small firms was not in keeping with our expectations. Unlike a number of the studies we have discussed, our tests have also taken care to explicitly address the errors-in-variables problem. We believe that there are some important messages in our study. It is clearly not the case that beta-based models are simply bad (as the mainstream of the literature might suggest): the results for our analyses of the orthogonalized hybrid model indicate that such a conclusion is too strong. The findings in this study suggest that it is important to consider asymmetry between up and down markets in asset pricing tests. Our study also provides evidence for the potential importance of considering world systematic risk in capturing the cross-section of returns. Furthermore, it is clear that asset pricing tests need to consider the information sets used by investors to condition their expectations. Finally, different models are appropriate for different countries. Our study therefore provides a further reminder that a “one size fits all” approach to asset pricing may lead to invalid inferences about investors' expectations of returns and the cost of capital.