range, the easier it is for the manager to stay within it. Also, with a larger target range or green zone, a departure from the range clearly represents a stronger signal. Thus, there will always be a tension that needs to be balanced in setting the size of the target range. These issues in the target setting process need to be recognized when budgeting active risk at the asset class and plan levels as well. Obviously, if we give managers a longer tracking error leash, it impacts our ability to manage targeted levels of asset class and plan risk. These issues will be highlighted later in our discussion. When determining the exact boundaries for the targeted green, yellow, and red zones, we would suggest using the following framework. Earlier, we defined the yellow zone as an "unsuccessful" outcome. Unsuccessful in this context is somewhat arbitrary, however, so we suggest defining it as something that in normal markets should be expected to happen no more than one or two times per year on either the downside or the upside. Put another way, we would expect to set the targeted green zone wide enough such that it would cause the realized tracking error to exit the targeted zone no more than twice per year. We also defined the red zone to be a set of "bad" or "rare" outcomes. Again, in a somewhat arbitrary fashion, however, we can build the red zones by appropriately setting the upper and lower boundaries for our yellow zones. In the case of the red zone, we define "rare" as an event that goes beyond the yellow zone on the upside or the downside no more than one or two times in five years. It is important to note that while we are introducing a relatively simple color-coded approach to managing tracking error, we also recognize that the simplicity of this approach may be deceiving. The random influences of environmental factors in different markets, as well as the complexities of portfolio construction, statistical estimation, and so on, lead quickly to a thicket of complicated issues when one attempts to apply this approach in practice. Nonetheless, as yellow and red warnings occur, such issues are very relevant to the risk manager in interpreting the cause and implications of the signal. While the use of either daily or monthly data in the target-setting process is appropriate, we would strongly advocate the use of daily data for tracking error computations after managers are hired for an assignment. Daily performance data coming in the form of manager feeds or performance attribution systems will help investors in identifying tracking error issues before they impact performance. Understandably, rolling 20-day and 60-day tracking error estimates can at times be noisy, yet they provide a reasonably accurate depiction of what is going on in the portfolio at the time. Therefore, we believe that shorter estimation periods can also be a leading indicator and highlight potential issues in a manager's portfolio. Finding out relatively quickly allows a risk manager to react equally as fast. For example, if our international equity manager's targeted tracking error is 550 to 1,000 basis points and we compute the most recent 60-day tracking error of his portfolio to be 400 basis points, then clearly this is an indication to the risk manager that further analysis is required to better understand the associated exposures that are leading to the unexpectedly low tracking error. Low tracking error is of as much concern as high tracking error because it makes achieving return targets more difficult. If we were constrained by the frequency of monthly data for our risk analysis,