Capital Budgeting: IT Project Portfolio Optimization Redux
Probability and Statistics
In general, risk or uncertainty refers to the probability that some unfavorable event will occur. A probability distribution is defined as a set of possible outcomes, with a probability of occurrence attached to each outcome. The values for resource (Manpower and Capital) usage in Figure 1 appear to be constants. But, under the hood, they are probability distributions such as those shown in Figure 2.
Figure 3: Specify correlations between your inputs via correlation matrices.
During a simulation, it is important to account for any correlation between input variables. Correlation occurs when the sampling of two or more input distributions are related. But, the random numbers generated by the pop-up windows (probability distributions) in Figure 2 are probabilistically independent. This means, for example, that if a random value in one cell is much larger than its mean, the random values in other cells are completely unaffected. They are no more likely to be abnormally large or small than if the first value had been average or less than average. Sometimes, however, this independence is unrealistic. Therefore, you want the random numbers to be correlated in some way. If they are positively correlated, large numbers tend to go with large numbers, and small with small. If they are negatively correlated, then large tend to go with small and small with large.
The dialog window shown in Figure 3 allows you to specify correlations between your inputs through correlation matrices. This limited illustration specifies a positive correlation of .8 between available capital and available resources for Project A. This might be the case for a project in which the cost of human resources generally consumed a very large part of the investment.
Figure 4: Specifying the goals, variables, and limits used by the optimizer.
Figure 4 shows the main dialog box by which you link the RISKOptimizer optimizer to the Excel spreadsheet. The dropdown list in Figure 4 is shown open in Figure 5.
Figure 5: Selecting the statistic to maximize or minimize
To select the statistic for the target cell which you wish to minimize, maximize, or set to a specific value, simply select the desired statistic from the displayed dropdown list. If you want to select a Percentile or Target for the target cell's distribution, simply select Percentile(x) or Target(x), where "x" is an actual value.
Politics, Biases, and Other Vulnerabilities
For internal political reasons, management on occasion will be prepared to rearrange project priorities. In a sense, the capital budgeting decision represents a resolution of conflict (hopefully not too unfriendly) among colleagues in the same organization. In reality, the management of each division or department is constantly trying to outmaneuver its internal competition within the firm to win a bigger slice of the available funds. In such a situation, political and prestige factors sometimes may overrule rational decision-making, but in the longer run the race is not always to the swift, and the ability to present cogent arguments based on careful and accurate forecasting of the relevant cashflows is crucial.
At the same time that cash flow estimation is the most critical part of the capital budgeting process, it's also the most difficult. Unfortunately, several studies indicate that capital budgeting cash flow forecasts are not unbiased—rather, managers tend to be overly optimistic in their forecasts, and as a result, revenues tend to be overstated and costs tend to be understated. The end result is an upward bias in net operating cash flows and thus an upward bias in estimated NPVs.
Forecasting errors result from our personal biases: For example, supporting evidence bias is our tendency to want to confirm what we already suspect and look for facts that support it. We avoid asking tough questions and discount new information that might challenge our preconception. Suppose, for example, you are considering an investment to automate some internal business function. Your first inclination is to call an acquaintance who has been boasting about the good results his or her organization obtained from doing the same. What response do you expect other than, "It's the right choice"?
Despite an inclination to look for supporting evidence, it is usually much more informative to seek out contradictory evidence. Confirming evidence often fails to discriminate among possibilities well. To illustrate, in one study students were given the initial sequence of numbers 2, 4, 6 and told to determine the rule that generated the numbers. To check hypotheses, they could choose a possible next number and ask whether that number was consistent with the rule. Most students asked whether a next number "8" would be consistent with the rule. When told it was, they expressed confidence that the rule was, "The numbers increase by 2." Actually, the rule was, "Any increasing sequence." A better test would have been to check whether a next number incompatible with their hypothesis (for example, "7") was consistent with the unknown rule.
Sometimes, poor forecasting occurs because divisional (or departmental) managers are rewarded on the basis on the size of their divisions (or departments); hence, they are motivated to maximize the number of projects accepted rather than the profitability of the projects. Even when this is not the case, managers often become emotionally attached to their projects, and therefore are unable to objectively assess a project's potential negative factors.
A first step in recognizing poor cash flow estimation, especially for projects that are estimated to be highly profitable, is to ask this question: What is the underlying cause of this project's profitability? If the firm has some inherent advantage, such as patent protection, unique expertise, or even a well-regarded brand name, projects that utilize such an advantage may truly be extraordinarily profitable. However, in the longer run, profitability will probably be eroded by competition until the returns on projects within an industry are close to the normal return, which is the cost of capital.
If there is reason to believe that this situation exists, and if division (department) managers cannot identify any unique factors that could support a project's estimated high profitability, senior management should be concerned about the possibility of estimation bias. Recognizing that biases may exist, senior managers at many firms develop data on the forecast accuracies of their divisional (departmental) managers, and then consider this information in the capital budgeting decision process. Some companies lower the cash flow estimates of managers whose track records suggest that their forecasts are too rosy, while other companies increase the cost of capital, or hurdle rate, applied to such project submissions. I'll have more to say on the cost of capital in the next article of this series.
The computational side of complex project portfolio selection is greatly aided by the availability of relative low-cost, high-speed computers on which you can model the complex equations of modern economic theory. However, this high technology is only one link in the chain of things all of which must be done right before you can deliver a beneficial capital budget. So, before you do anything else, you need to address the issues outlined in the "Politics, Biases, and Other Vulnerabilities" section. Otherwise, the good names of Mathematics, Economics, and the like may be besmirched by the bad practice of more "Garbage in, Garbage out."
Page 2 of 3