factor or asset. If there is a systematic source of residual, then we expect this to be picked up by a daily monitoring process that measures and evaluates the one-day residual returns for each portfolio that is tracked. This daily monitoring process increases the likelihood that systematic sources of residuals are identified in a timely manner. To better understand the impact that a residual can have on return attribution, suppose we are interested in computing return attribution on a portfolio over a six-month period (126 business days) and, each business day, the residual is 0.25 basis points. If we ran a one-day attribution on any day over the period, the residual would be too small to see since our reports show numbers in whole basis points and not fractions. However, assuming that the residual is constant over the period, the six-month compounded portfolio return would have a residual of about 32 basis points (126 0.25 bps). In order to reduce the six-month residual, we apply the adjustment algorithm described, each day, to the sources of return. If we had 100 assets (sources of return) in the active portfolio, then we would be modifying the contribution of each asset by a maximum of .25/100 bps or 0.0025 bps per day. The compounded adjustment to each source of return over the six-month period is, on average, 0.32 basis points. Moreover, the original ranking of the sources is unaffected. An algorithm such as the one described should be applied only if the magnitude of the residual is considered small enough as to not materially affect the results. Typically, it requires that we have daily, officially reported returns. Without the official returns, the algorithm cannot be applied. Finally, we present an additional reason for computing the residual as frequently as possible. Suppose that a manager has a return attribution report and the residual on the managed portfolio's return for the particular month is 0.5 bps. The manager of an equity portfolio might view this error as small, particularly if the return on the portfolio is relatively big-say, 5 percent. The question that we pose is, is the error really small? To answer this question, a manager might look at each day's residual during the month-that is, taking daily position files, compute the difference between the managed portfolio's return and the official return, each day, over the attribution period. Suppose the manager finds that each day's residual is negligible, except for two days out of the month. On those days, the residuals are 50 bps and -51 bps. Since the sum of the daily residuals is approximately equal to the monthly residual, we might feel uncomfortable concluding that the monthly residual is small. In fact, the monthly residual may very well be meaningless. SUMMARY Return attribution is the process in which sources of a portfolio's return are identified and measured. Managers may rely on return attribution reports developed in-house or from commercially available systems. Differences across systems can vary in certain respects, from the algorithms applied to the terminology used to describe