achieve the median information ratio as the second manager. From a practical perspective, if we truly believed that we could not differentiate between managers, then the lower confidence on manager 1 and higher confidence on manager 3 is indicating that we should reallocate risk away from manager 3 and into manager 1. Now let's look at the case when we have views that the information ratios differ by manager. In this example, the confidence levels are the same across managers. More specifically, in this example we are confident in the ability of each manager to hit its respective expected alphas. Separating out the impact of confidence levels and views is an important step to take in understanding the risk budget. Just as in the previous example, where we allocated active risk at the asset class level (e.g., U.S. Large Cap equity versus Core+ Fixed Income), we can start to identify factors that affect views, and those that affect confidence levels at the individual manager level. For example, suppose that we take as our starting view that each manager in an asset class will earn the median information ratio. We would change that view for a particular manager if there were structural factors that made us believe that they could outperform the median. An example would be the impact of no net short constraints: Lower tracking error managers are usually less susceptible to these constraints, suggesting that their expected information ratios should be higher. What would influence our choice of confidence in one manager versus another? One factor that we could consider is the length of the track record. All else being equal, we might be more confident in managers with longer track records than those with shorter histories. We might then believe that more risk should be allocated to those managers with longer track records. A second factor that might influence our confidence in one manager versus another is the stability of the team. Investment managers with less stable teams might cause us to dampen the degree of confidence, and consequently take more risk with other, more stable teams. Third, we might consider the risk "footprint" of one manager versus another. Consider, for example, two managers with the same information ratios and historical tracking errors. However, suppose that one manager seems to switch (for no apparent reason) between low and high tracking error regimes, while the other does not. Because the reasons for the switch between regimes are not evident, we might be less confident in the first manager. CONCLUSIONS Developing an allocation of active risk is an important part of the design of any investment policy. The allocation of active risk across strategies sets the framework for the ongoing evaluation of specific active strategies and specific investment managers. In this chapter, we have illustrated how active risk budgeting can be used to approach this issue. The predictions of asset pricing theory are quite clear about the return to active risk: in equilibrium it is zero. Nonetheless, because active risk is uncorrelated with market risk and because markets over the short term are not in equilibrium, in-