Raymond Kan

Optimal Portfolio Choice with Parameter Uncertainty

In this paper, we analytically derive the expected loss function associated with using sample means and the covariance matrix of returns to estimate the optimal portfolio. Our analytical results show that the standard plug-in approach that replaces the population parameters by their sample estimates can lead to very poor out-of-sample performance. We further show that with parameter uncertainty, holding the sample tangency portfolio and the riskless asset is never optimal. An investor can benefit by holding some other risky portfolios that help reduce the estimation risk. In particular, we show that a portfolio that optimally combines the riskless asset, the sample tangency portfolio, and the sample global minimum-variance portfolio dominates a portfolio with just the riskless asset and the sample tangency portfolio, suggesting that the presence of estimation risk completely alters the theoretical recommendation of a two-fund portfolio.

Two‐Pass Tests of Asset Pricing Models with Useless Factors

In this paper we investigate the properties of the standard two-pass methodology of testing beta pricing models with misspecified factors. In a setting where a factor is useless, defined as being independent of all the asset returns, we provide theoretical results and simulation evidence that the second-pass cross-sectional regression tends to find the beta risk of the useless factor priced more often than it should. More surprisingly, this misspecification bias exacerbates when the number of time series observations increases. Possible ways of detecting useless factors are also examined. Copyright The American Finance Association 1999.

Tests of Mean-Variance Spanning

In this paper, we conduct a comprehensive study of tests for mean-variance spanning. Under the regression framework of Huberman and Kandel (1987), we provide geometric interpretations not only for the popular likelihood ratio test, but also for two new spanning tests based on the Wald and Lagrange multiplier principles. Under normality assumption, we present the exact distributions of the three tests, analyze their power comprehensively. We find that the power is most driven by the difference of the global minimum-variance portfolios of the two minimum-variance frontiers, and it does not always align well with the economic significance. As an alternative, we provide a step-down test to allow better assessment of the power. Under general distributional assumptions, we provide a new spanning test based on the generalized method of moments (GMM), and evaluate its performance along with other GMM tests by simulation.

A Critique of the Stochastic Discount Factor Methodology

In this paper, we point out that the widely used stochastic discount factor (SDF) methodology ignores a fully specified model for asset returns. As a result, it suffers from two potential problems when asset returns follow a linear factor model. The first problem is that the risk premium estimate from the SDF methodology is unreliable. The second problem is that the specification test under the SDF methodology has very low power in detecting misspecified models. Traditional methodologies typically incorporate a fully specified model for asset returns, and they can perform substantially better than the SDF methodology. Copyright The American Finance Association 1999.

The Distribution of the Sample Minimum-Variance Frontier

In this paper, we present a finite sample analysis of the sample minimum-variance frontier under the assumption that the returns are independent and multivariate normally distributed. We show that the sample minimum-variance frontier is a highly biased estimator of the population frontier, and we propose an improved estimator of the population frontier. In addition, we provide the exact distribution of the out-of-sample mean and variance of sample minimum-variance portfolios. This allows us to understand the impact of estimation error on the performance of in-sample optimal portfolios.

Specification Tests of Asset Pricing Models Using Excess Returns

In this paper, we discuss the impact of different formulations of asset pricing models on the outcome of specification tests that are performed using excess returns. We point out that the popular way of specifying the stochastic discount factor (SDF) as a linear function of the factors is problematic because (1) the specification test statistic is not invariant to an affine transformation of the factors, and (2) the SDFs of competing models can have very different means. In contrast, an alternative specification that defines the SDF as a linear function of the de-meaned factors is free from these two problems and is more appropriate for model comparison. In addition, we suggest that a modification of the traditional Hansen-Jagannathan distance (HJ-distance) is needed when we use the de-meaned factors. The modified HJ-distance uses the inverse of the covariance matrix (instead of the second moment matrix) of excess returns as the weighting matrix to aggregate pricing errors. Asymptotic distributions of the modified HJ-distance and of the traditional HJ-distance based on the de-meaned SDF under correctly specified and misspecified models are provided. Finally, we propose a simple methodology for computing the standard errors of the estimated SDF parameters that are robust to model misspecification. We show that failure to take model misspecification into account is likely to understate the standard errors of the estimates of the SDF parameters and lead us to erroneously conclude that certain factors are priced.

Modeling non-normality using multivariate t: implications for asset pricing

Purpose - The purpose of this paper is to show that multivariate Design/methodology/approach - The EM algorithm is applied to solve the statistical estimation problem almost analytically, and the asymptotic theory is provided for inference. Findings - The authors find that the multivariate normality assumption is almost always rejected by real stock return data, while the multivariate Practical implications - The results provide improved estimates of cost of capital and asset moment parameters that are useful for corporate project evaluation and portfolio management. Originality/value - The authors proposed new procedures that makes it easy to use a multivariate

Misspecification-Robust Inference in Linear Asset-Pricing Models with Irrelevant Risk Factors

We show that in misspecified models with useless factors (for example, factors that are independent of the returns on the test assets), the standard inference procedures tend to erroneously conclude, with high probability, that these irrelevant factors are priced and the restrictions of the model hold. Our proposed model selection procedure, which is robust to useless factors and potential model misspecification, restores the standard inference and proves to be effective in eliminating factors that do not improve the models pricing ability. The practical relevance of our analysis is illustrated using simulations and empirical applications.

A New Variance Bound on the Stochastic Discount Factor

In this paper, we construct a new variance bound on any stochastic discount factor (SDF) of the form m = m(x), with x being a vector of state variables, which tightens the well-known Hansen-Jagannathan bound by a ratio of one over the multiple correlation coefficient between x and the standard minimum variance SDF, m0. In many applications, the correlation is small, and hence the bound is much improved. For example, when x is the growth rate of consumption, the new variance bound can be 25 times greater than the Hansen-Jagannathan bound, making it much more difficult to explain the equity-premium puzzle.

On the distribution of the sample autocorrelation coefficients

Sample autocorrelation coefficients are widely used to test the randomness of a time series. Despite its unsatisfactory performance, the asymptotic normal distribution is often used to approximate the distribution of the sample autocorrelation coefficients. This is mainly due to the lack of an efficient approach in obtaining the exact distribution of sample autocorrelation coefficients. In this paper, we provide an efficient algorithm for evaluating the exact distribution of the sample autocorrelation coefficients. Under the multivariate elliptical distribution assumption, the exact distribution as well as exact moments and joint moments of sample autocorrelation coefficients are presented. In addition, the exact mean and variance of various autocorrelation-based tests are provided. Actual size properties of the Box-Pierce and Ljung-Box tests are investigated, and they are shown to be poor when the number of lags is moderately large relative to the sample size. Using the exact mean and variance of the Box-Pierce test statistic, we propose an adjusted Box-Pierce test that has a far superior size property than the traditional Box-Pierce and Ljung-Box tests.