is also very important in determining the appropriate mix of strategies at the portfolio level. Again, a mix of historical results and intuition about the future can help determine these correlation estimates. For example, it makes sense that security selection strategies should not exhibit much correlation at all with other strategies because the manager's top-down view does not necessarily factor into individual security opinions. On the other hand, it might be expected that the duration strategy would have some positive correlation with the sector allocation strategy because the manager's macroeconomic opinion will be an input into both decisions. Step 6 Determine the target excess return or the target tracking error. This is a necessary input, as the allocation of risk will seek either to maximize the excess return given a target level of risk or to minimize the tracking error needed to achieve a target level of excess return. Some clients give one target or the other, some give both, and some give neither. In the first two cases, the manager must make sure that the targets are achievable given the constraints. In cases where clients give no explicit risk or return targets, managers must use their judgment given their understanding of the clients' objectives. Step 7 Determine the optimal amount of risk to each strategy. Finally, once you determine what strategies can be used, how much could be allocated to each strategy, how good you are at generating returns in each strategy, how you expect them to move together, and how much risk or return you are looking to achieve, you are ready to determine the appropriate allocation. We use a mean-variance optimization technique that utilizes all of the above inputs and results in an allocation of tracking error to each strategy that will maximize the information ratio. In order to demonstrate this process, we will perform a few examples of this optimization technique. For purposes of this illustration, we will use several inputs: 1.    The portfolios are managed to the Lehman Brothers U.S. Aggregate Index. 2.    The portfolios are long only and do not allow explicit leverage. 3.    The estimated information ratios for the active strategies are: Duration 0.2 Yield curve 0.3 Sector allocation 0.4 Security selection for government/MBS/ABS/agency 0.7 Security selection for corporate/high-yield/EMD 0.5 Country allocation 0.4 Currency 0.4 4. The correlations of the strategies range from 0.0 to 0.3. We have run the optimization for two different sets of constraints (see Table 24.4). Table 24.5 shows the output from the optimization for Portfolio 1 where risk was allocated across the allowable strategies with three different targets for tracking error of 50, 75, and 100 basis points.