We combine work in investments and asset pricing because of their obvious synergies.

Devraj Basu and Alex Stremme work on asset pricing, making use of the modern techniques of stochastic discount factors.  Stewart Hodges works on a number of normative problems in fund management in collaboration with some past and present PhD students.  He also maintains contacts with a number of investment professionals. 

Current research topics include:

Investments:

• Correcting for data-mining biases in portfolio optimisation,

• The value of fundamental information,

• Forecasting the risk/tracking error on equity portfolios.

Asset Pricing:

• the optimal use of predictive variables in forming portfolios and the dual issue of using
   these to test of asset pricing models,

• conditional performance evaluation.

Investment Management

The necessity of scaling forecasts for portfolio optimisation purposes was first pointed out in an early paper by Hodges and Brealey (Financial Analysts Journal, 1972).  This led to the terminology “Information Coefficient” as a more marketable synonym for correlation after the approach was adapted as a commercial product by Keith Ambachsheer at Canavest House.  Remarkably, the issue is still of current interest, particularly in applications to dynamic asset allocation.  Stewart Hodges is collaborating with Nato Kemkhadze in this area.

Another current project considers the value of having good information about fundamentals, when asset prices continue to deviate from these values.  We see this very directly in closed-end investment trusts, where the purchase of a fund at a discount to its asset value is by no means a guarantee of superior investment performance.  The current work uses a continuous time model to explore how “right” you have to be to obtain significant economic benefits from trading.  Stewart Hodges is collaborating with Jennifer Liao at Manchester.

Managed funds are increasingly measured against formal benchmarks and need to be able to assess their future tracking error as accurately as possible.  Stewart Hodges is working with Lan Lin on alternative methods for forecasting risk and tracking error on portfolios of equities. 

Asset Pricing

A major goal of asset pricing is to understand the prices of financial assets such as stocks, bonds and derivatives. An important issue here is explaining the cross section of stock and bond returns. Recent research seems to indicate that there is predictability in short and long horizon stock returns and that commonly used macro-economic as well as firm level variables have predictive power. The nature and interpretation of this return predictability continues to be debated, but there is an impressive body of recent evidence that points to its existence. Our recent research has focused on the implications of predictability for asset pricing models and asset allocation.

The first issue that we have tackled is the optimal use of these predictive variables in forming portfolios and the dual issue of optimally incorporating these variables in tests of asset pricing models. We provide a unified approach to these issues by implementing the Hansen and Richard (1987) approach to incorporating conditioning information in asset pricing models. Our approach to incorporating conditioning information in asset pricing models. Our approach yields a unified derivation of Hansen-Jagannathan type bounds which allows us to compare and contrast the bounds due to Ferson and Siegel (2003) and Bekaert and Liu (2004). Focusing on the asset allocation we provide a method to assess the economic value of return predictability, both in-sample and out of sample.

Another major theme in current asset pricing research is studying asset pricing models with time varying risk premium. This is motivated by the inability of the static CAPM to explain the cross section of stock returns. Several studies have focused on models where the factor loadings or betas are functions of macro-economic or firm-level variables. A current research project focuses on implementing the Hansen and Richard approach in this context to derive the ‘optimal’ factor loadings. The next issue we plan to tackle is conditional performance evaluation, an issue of considerable practical significance.