In the analysis of portfolio management, the initial work of Markowitz (1959) was directed towards finding the optimal weights in a portfolio. It was quickly realized that the decisions involved in building up a portfolio were less frequent than the decisions to modify existing portfolios. This is especially important when analyzing how profitable portfolio managers have been over time. If, for example, a portfolio consists of equities and bonds, some investment managers might be particularly ski lled in choosing specific companies in which the portfolio should invest, while others might be able to forecast at which times the portfolio should be more heavily invested in shares. The first type of skill would be classified as being more concerned with portfolio selection while the latter would be described as connected with timing or asset allocation.
Asset allocation decisions can be further divided. Investors can decide on an ad hoc basis to alter their portfolio by changing the weights of the constituent assets as a result of some specific model. For example, forecasting models are used to predict the performance of equities relative to bonds or real estate relative to equities. Dependent on the outcome of these forecasts, the investor will switch into or out of the asset being forecasted. Models are used to derive frequent forecasts of one asset against another and to move the portfolio day by day depending on the outcome of the forecasting model. This type of model is sometimes referred to as tactical asset allocation (TAA) and in practice is used in conjunction with some sophisticated trading in derivatives such as options or futures. Instead of buying more shares, this system buys options or futures in an index representing equities. If equities rise in value, so will the options and futures position and the portfolio thereby will increase in value to a greater extent than underlying equities. TAA i s used to adjust portfolio exposure to various factors such as interest rates and currency movements as well as overseas investments (see Arnott et al., 1989).
An alternative category of asset allocation is the technique of dynamic asset allocation, where there is less emphasis on forecasting which component assets will perform well in the next period and more on setting up a policy by which the portfolio reacts automatically to market movements. This can be organized with the help of options and futures but can also be carried out by adjusting the weights of the component assets in the light of predetermined rules. For example, the policy of buying an as set when that asset has performed well in the current period and selling when it has done badly can be carried out in such a way as to provide portfolio insurance, i.e. it protects the portfolio by reducing the exposure to successive falls in the value of one of its constituent assets. An alternative dynamic asset allocation policy is that carried out by rebalancing so as to maintain a reasonably constant proportion in each asset. This involves selling those assets which have just risen in value and selling those assets which have just fallen in value. The two strategies are profitable in different phases of the market. When the market is moving strongly, the insurance policy is most successful. If, however, the market is tending to oscillate without a strong trend, the rebalancing policy works best. These principles are well illustrated in Perold and Sharpe (1988).
