Mental biases are not our only bias. Another kind of bias arises when one person is acting on behalf of another. This is called an agency problem—a situation in which the owner of a project has to rely on someone else for information, and this someone else has divergent interests. An example may be shareholders who rely on corporate management to undertake projects on their behalves, or a division manager who has to rely on department managers for information about how profitable their proposed projects really are. A cynical synopsis of agency biases would be “all people act and lie in their own self-interest.” Now, although everyone does have incentives to lie—or at least color the truth—corporations are especially rife with such agency distortions. Of course, few people sit down and contemplate how to best and intentionally lie. Instead, they convince themselves that what is in their best interest is indeed the best route to take. Thus, mental biases often reinforce incentive problems: “wishful thinking” is a disease from which we all suffer.
You can take the fact that we have already had to mention agency issues repeatedly in this blog as an indication of how important and pervasive these are. But, again, lack of space forces us to highlight just a few issues with some examples:
1. Competition for Capital Managers often compete for scarce resources. For example, division managers may want to obtain capital for their projects. A less optimistic but more accurate estimate of the project cash flows may induce headquarters to allocate capital to another division instead. Thus, division managers often end up in a race to make their potential projects appear in the most favorable and profitable light.
2. Employment Concerns Managers and employees do not want to lose their jobs. For example, scientists tend to highlight the potential and downplay the drawbacks of their areas of research. After all, not doing so may cut the project and thereby cost them their jobs.
3. Perks Managers do not like to give up perks. For example, division managers may like to have their own secretaries or even request private airplanes. Thus, they are likely to overstate the usefulness of the project “administrative assistance” or “private plane transportation.”
4. Power Managers typically love to build their own little “empires.” For example, they may want to grow and control their department because bigger departments convey more prestige and because they are a stepping stone to further promotion, either internally or externally. For the same reason, managers often prefer not to maximize profits, but sales.
5. Hidden Slack Managers like the ability to be able to cover up problems that may arise in the future. For example, division managers may want to hide the profitability of their divisions, fearing that headquarters may siphon off “their” profits into other divisions. They may prefer to hide the generated value, feeling that the cash they produced in good times “belongs” to them—and that they are entitled to use it in bad times.
6. Reluctance to Take Risk Managers may hesitate to take on risk. For example, they may not want to take a profitable NPV project, because they can only get fired if it fails—and may not be rewarded enough if it succeeds. A popular saying used to be “no one was ever fired for buying IBM,” although these days Microsoft has taken over IBM’s role.
7. Direct Theft Managers and employees have even been known to steal outright from the company. For example, a night club manager may not ring sales into the cash register. Or a sales agent may “forget” to charge her relatives. In some marginal cases, this can be a fine line. For example, is taking a paper clip from the company or answering a personal e-mail from the company account really theft? In other cases, theft is blatantly obvious. In September 2002, Dennis Kozlowski, former CEO, was charged with looting $600 million from Tyco shareholders. His primary defense was that he did so in broad daylight—with approval from the corporate board that he had helped put in place.

Most cash flow and cost-of-capital estimates rely on judgments. Unfortunately, it is often difficult to obtain accurate judgments. Our brains tend to commit systematic decision errors.
Managers who do not recognize these biases will systematically make poor decisions.
There are literally dozens of well-known behavioral errors, but limited space allows us to highlight just three: overconfidence, relativism, and compartmentalization.
1. Overconfidence is the tendency of people to believe that their own assessments are more accurate than they really are. In lab experiments, ordinary people are found to be dramatically overconfident. When asked to provide a 90% confidence interval—which is just a range within which they are confident that their true value will lie in nine out of ten tries—most people end up being correct only five out of ten times.
It is difficult to empirically document overconfidence—after all, if it were easy, managers would recognize it themselves and avoid it. However, we do have evidence that many managers who are already heavily invested in their own company tend to throw caution overboard and voluntarily invest much of their own money into the corporation—and even in companies going bankrupt later on. There is also good empirical evidence that those of us who are most optimistic in overestimating our own life-expectancy disproportionately become entrepreneurs. Even if optimism is a disease, it seems to be a necessary one for entrepreneurs!
To understand this better and to test your own susceptibility to these problems, you can take a self-test at the website, http://www.cashtrailer.com/. Doing so will likely make you remember this problem far more than reading long paragraphs of text in this blog. Incidentally, the only population segments who are known not to be systematically overconfident are weather forecasters and clinically depressed patients.
2. Relativism is the tendency of people to consider issues of relative scale when they should not. For example, most people are willing to drive 15 minutes to a store farther away to save $20 on the purchase of $30 worth of groceries, but they would not be willing to drive the 15 minutes to a car dealer farther away to save $100 on the purchase of a new $20,000 car. The savings appears to be less important in the context of the car purchase than in the context of a grocery purchase. This is flawed logic, similar to comparing IRR’s while ignoring project scale. The marginal cost is driving 15 minutes extra, and the marginal benefit is a higher $100 in the context of the car than the $20 in the context of the groceries. Put differently, the problem is that we tend to think in percentages, and the $20 is a higher percentage of your grocery bill than it is of your car purchase. The smaller the amount of money at stake, the more severe this problem often becomes. When a gas station advertises a price of $2 per gallon rather than $2.10, customers often drive for miles and wait in long lines—all to fill a 20 gallon gas tank at a total savings that amounts to a mere $2.
3. Compartmentalization Compartmentalization is the tendency of people to categorize decisions. Most people are more inclined to spend more when the same category has produced an unexpected windfall earlier. For example, winning a lottery prize while attending a baseball game often makes winners more likely to purchase more baseball tickets, even though the project “baseball game” has not changed in profitability. Similarly, an unexpected loss may stop people from an otherwise profitable investment that they should make. For example, say an individual likes to attend a particular baseball game. If she loses her baseball game ticket, she is less likely to purchase a replacement, even though the cost and benefit of purchasing the ticket are the same as they were when the original ticket was purchased.

But should we really charge zippo for parking corporate cars if we suspect that the unused capacity will not be unused forever? What if a new division might come along that wants to rent the five currently unused garage space in the future? Do we then kick out all current parkers? Or, how should we charge this new division if it wanted to rent six spaces? Should we give it the five remaining unused parking spots for free? Presuming that garages can only be built in increments of ten parking spots each, should we build another ten-car garage, and charge it entirely to this new division that needs only one extra parking spot in the new garage? Should this new division get a refund if other divisions were to want to use the parking space— but, as otherwise unused parking space, should we not use the garage appropriately by not charging for the nine extra spaces that will then be a free resource?
When there are high fixed and low variable costs, then capacity is often either incredibly cheap (or even free) or it is incredibly expensive—at least in the short run. Still, the right way to think of capacity is in terms of the relevant marginal costs and marginal benefits. From an overall corporate perspective, it does not matter how or who you charge—just as long as you get the optimal capacity utilization. To the extent that cost allocation distorts optimal marginal decision-making, it should be avoided. In our case, if optimal capacity utilization requires zero parking cost for the old garage, then so be it. Of course, when it comes to the decision to build an entirely new garage, you simply weigh the cost of building the 10-spot garage against the reduced deterioration for 1 car.
Unfortunately, real life is not always so simple. We know that our division managers will not want to pay for it if they can enjoy it for free—so we cannot rely on them telling us the correct marginal benefit. So, would it solve our problem to charge only divisions that are voluntarily signing up for the Internet connection, and to forcibly exclude those that do not sign up? If we do, then we solve the problem of everyone claiming that they do not need the Internet connection. However, we are then stuck with the problem that we may have a lot of unused network capacity that sits around, has zero marginal cost, and could be handed to the non-requesters at a zero cost. It would not impose a cost on anyone else and create more profit for the firm. Of course, if we do this, or even if we are suspected to do this, then no division would claim that they need the Internet to begin with, so that they will get it for free. In sum, what makes these problems so difficult in the real world is that as the boss, you often do not know the right marginal benefits and marginal costs, and you end up having to “play games” with your division managers to try to make the right decision. Such is real life!

A closely related mistake is to forget that “overhead” is often a sunk cost. By definition, over-head is not a marginal cost, but something that has been incurred already and is allocated to departments. For example, assume your firm has spent $500,000 on a computer that is currently idle half the time. It serves only one division. Assume that another division can take an additional project that produces $60,000 in net present value, but that will consume twenty percent of the computer’s time. Should your firm take this project? If twenty percent of the cost of the computer is allocated to this new project (i.e., 20% · $500, 000 = $100, 000), the net present value of the new project would appear to be a negative −$40, 000. But the correct decision process is not to allocate the existing overhead as a cost to divisions. The $500,000 on overhead has already been spent. The computer is a sunk cost—assuming that it really would sit idle otherwise and find no better purpose. It may seem unfair to have charged only the original division for the computer and exempt the opportunistic other division. Yet taking this additional project will produce $60,000 in profits without cost—clearly, a good thing. I personally know of plenty of examples in which overhead allocation has killed very profitable projects.
“Capacity” is a subject that is closely related. For example, a garage may be currently only used for half its space. Adding the project “another car” that could also park in the garage would reduce this car’s depreciation. The garage would then have a positive externality on project “corporate cars.” The marginal cost of storing other cars in the garage should be zero.

Sunk costs are, in a sense, the opposite of marginal costs. A sunk cost is a cost that cannot be altered and that therefore should not enter into your decisions today. It is what it is. Sunk costs are ubiquitous, if only because with the passage of time, everything is past or irrevocably decided and thus becomes a sunk cost.
For example, consider circuit board production—a very competitive industry. If you have just completed a circuit board factory for $1 billion, it is a sunk cost. What matters now is not that you spent $1 billion, but how much the production of each circuit board costs. Having invested $1 billion is irrelevant. What remains relevant is that the presence of the factory makes the marginal cost of production of circuit boards very cheap. It is only this marginal cost that matters when you decide whether to produce circuit boards or not. If the marginal board production cost is $100 each, but you can only sell them for $90 each, then you should not build boards, regardless of how much you spent on the factory. Though tempting, the logic of “we have spent $1 billion, so we may as well put it to use” is just plain wrong. Now, presume that the market price for boards is $180, so you go ahead and manufacture 1 million boards at a cost of $100 each. Alas, your production run has just finished, and the price of boards—contrary to everyone’s best expectations—has dropped from $180 each to $10 each. At this point, the board production cost is sunk, too. Whether the boards cost you $100 to manufacture or $1 to manufacture is irrelevant. The cost of the production run is sunk. If boards now sell at $10 each, assuming you cannot store them, you should sell them for $10 each. Virtually all supply costs eventually become sunk costs, and all that matters when you want to sell a completed product is the demand for the product.
One more note—time itself often, but not always, decides on what is sunk or not. Contracts may allow you to undo things that happened in the past (thereby converting an ex-post sunk cost into a cost about which you still can make decisions), or bind you irrevocably to things that will happen in the future.