The relationship between statistical measures of tracking error and asset allocation restrictions expressed as admissible weight ranges is discussed. Tracking errors are typically calculated as annualized second moments of return differentials between a portfolio and a benchmark. In practice, however, constraints on tactical deviations from benchmark weights are often imposed on the portfolio manager to ensure adequate tracking. Simulating various investment strategies subject to such constraints, how the size of acceptable deviations from the benchmark relates to the statistical tracking error is presented. An example based on actual market data indicates that imposing fairly large tactical asset allocation ranges produces surprisingly small tracking errors. It was also found that TAA restrictions should restrict not only the tactical ranges of the individual asset classes but also, and perhaps even more importantly, the tracking of individual asset classes.
The impact of the time horizon upon the risk of equity investments is still a controversial issue. In this paper, we analyse long-term risk in a portfolio insurance framework based on option pricing theory. The insurance strategies are implemented alternatively with a portfolio of stocks and put options or bonds and call options. The risk of stock holdings is measured by the permissible relative stock position in the replicating portfolio for an exogenous floor function. Our findings indicate that there is no general conclusion as to the long-term risk of stocks; the risk can only be determined for specific floor functions. Because the utility function is implicit in any floor specification, we argue that the assumption of preference-free determination of risk with the help of option-pricing theory, as recently suggested in the literature, is a fallacy. Moreover, the popular belief that a longer time horizon reduces the risk of equity investments and therefore makes it optimal to invest a greater fraction of one's wealth in stocks may not be justified.
In this study of tactical asset allocation, we use a genetic algorithm to implement a market timing strategy. The algorithm makes a daily decision whether to invest in the market index or in a riskless asset. The market index is represented by the S&P500 Composite Index, the riskless asset by a 3-month T-Bill. The decision of the genetic algorithm is based on fundamental macroeconomic variables. The association of fundamental variables with a set of operators creates a space of possible strategies from which the genetic algorithm attempts to select the optimal solution. To test its performance, we apply the genetic algorithm to different time periods of in-sample and out-of-sample data using rolling return estimates. In total, 39 different timing strategies are tested over the time period of 1980-2000. On a risk-adjusted basis, we observe a moderate outperformance for the timing strategy suggested by the algorithm compared to a passive index strategy. The forecasting power of the algorithm is higher during times of high volatility and pronounced changes in the return series. Moreover, the algorithm is more successful in forecasting long-term return patterns than short-term fluctuations.
We analyze the impact of prior performance on the risk-taking behavior of mutual fund managers. We contribute to the existing literature by using different measures of risks, a larger data set, and an econometric approach capturing non-linear effects and assigning exact probabilities to the mutual fund managers' adjustment of behavior. We find that prior performance in the first half of the year has, in general, a positive impact on the choice of the risk level in the second half of the year. Successful fund managers increase the volatility, the beta, and assign a higher proportion of their portfolio to value stocks, small firms, and momentum stocks in comparison to unsuccessful fund managers. Unsuccessful fund manager increase, on average, only the tracking error.
We use Markov Chain Monte Carlo (MCMC) methods for the parameter estimation and testing of conditional asset pricing models. In contrast to traditional approaches, it is truly conditional because the assumption that time variation in betas is driven by a set of conditioning variables is not necessary. Moreover, the approach has exact finite sample properties and accounts for errors-in-variables. Using S&P 500 panel data, we analyze the empirical performance of the CAPM and the Fama and French (1993) three-factor model. We find that time-variation of betas in the CAPM and the time variation of the coefficients for the size factor (SMB) and the distress factor (HML) in the three-factor model improve the empirical performance by a similar amount. Therefore, our findings are consistent with time variation of firm-specific exposure to market risk, systematic credit risk and systematic size effects. However, a Bayesian model comparison trading off goodness of fit and model complexity indicates that the conditional CAPM performs best, followed by the conditional three-factor model, the unconditional CAPM, and the unconditional three-factor model.
