The authors seek to determine whether “patience is a virtue” when considering the evaluation of managers hired to steward the assets of a particular plan. The authors use Monte Carlo simulation to generate lognormal returns for managers who exhibit a varying level of skill over time. The simulation includes three players: (1) the plan, which is an entity with financial assets that are invested in a portfolio of managers, (2) the chief investment officer (CIO), which is an algorithm that allocates plan assets to managers, and (3) the manager, which is an entity that can be hired to invest a share of plan assets.

Expected Sharpe ratios (i.e., alpha divided by standard deviation) of 1.0 (best), 0.75, 0.50, and 0.25 (worst) are implemented within the simulation. Because manager skill is known within the simulation, the subsequent performance of superior managers can be compared with that of lesser managers over time.

In the first set of simulations, managers of different skill levels are compared on the basis of the depth and duration of a drawdown within a three-year time frame. Although the most skilled managers are less likely to have a deep drawdown of long duration, the probability of such an event is not zero. When examining three-year periods for the best managers, two of the five best managers had a drawdown of at least 15 percent with a drawdown duration of about one year.

In a second set of simulations, 1,000 investable managers are simulated with 100 of the managers having an expected Sharpe ratio of 1.0 and the remaining 900 managers having an expected Sharpe ratio of 0. Three additional similar simulations are performed by changing the 100 skilled managers to have Sharpe ratios of 0.75, 0.50, and 0.25 for each respective simulation. Within each simulation, the 1,000 managers are ranked on the basis of performance during previous periods of varying length, which are referred to as the “ranking years.” As expected, the superior managers are more likely to have stronger subsequent performance, especially when longer ranking periods are used. Increasing the ranking years beyond five years, however, does not result in appreciable benefits.

What becomes evident is the consistent existence of “lucky” unskilled managers, and if the skilled managers are only “marginally” more skilled (e.g., a difference in the expected Sharpe ratio of 0.25), then the delineation of skilled versus unskilled becomes very difficult. Furthermore, the use of longer return histories results in smaller deviations between preselection and postselection returns. For example, using a three-year return history to select a portfolio of superior managers results in subsequent returns that are 52 percent lower than anticipated. Extending the preselection period to 10 years decreases the return difference to 10 percent.

In the next set of simulations, the authors examine the performance of CIO personalities based on (1) the length of the preselection period used to select managers, (2) the rule used for firing managers, and (3) whether managers are rehired. CIOs with longer-term perspectives are found to have better investment performance than those with shorter-term perspectives. The best CIO strategy is to use a five-year preselection period, fire managers if their performance falls into the lowest performance decile after one year, and rehire managers if their returns are in the top decile.

The authors conclude that the investment industry is currently too impatient. Investors should use longer evaluation periods to choose managers and be more patient in deciding whether to fire a manager. Exhibiting greater patience in manager hiring and firing decisions would result in a lower probability of hiring unskilled managers and less discrepancy between anticipated and realized returns.