Pitfalls in Evaluating Risky Projects James E. Hodder and Henry E. Riggs In recent years, the leaders of American companies have been barraged with attacks on their investment policies. Critics accuse American executives of shortsightedness and point out that managers in Japan and Europe often fix their vision on more dis¬tant horizons. Here, it is claimed, managers pay too much attention to quarterly earnings reports and not enough to such basic elements of industrial strength as research and development.

Some analysts see the root of this problem in the tendency of American companies to rely on discounted cash flow tech¬niques in weighing long-term investments. 1 These critics argue that DCF techniques have inherent weaknesses that make them inappropriate for evalu¬ating projects whose payoffs will come years down the road. We disagree with the contention that DCF tech¬niques are inappropriate for evaluating long-term or strategic investment proposals. We do believe, how- Mr. Hodder is assistant professor of industrial engineering and engineering management at Stanford University.

His teaching and research have focused on capital budgeting and international hedging decisions. Mr. Riggs is professor of industrial engineering and engineering management and vice president for development at Stanford University. Before joining the university in 1974, he worked for 15 years in industry in various financial positions. Riggs is the author of many publications on accounting and fi¬nance and of the recent book. Managing High-Technology Compa¬nies (Belmont, California, Lifetime Learnings Publications, a division of Wadsworth, Inc. , 1983). I. See for example, Robert H.

Hayes and David A. Garvin, “Manag¬ing as if Tomorrow Mattered,” HBR May-June 1982, p. 71. ever, that companies often misapply or misinterpret DCF techniques. Misuse is particularly serious in evaluating long-term capital investments, such as ambitious R&D projects, that appear to involve high risk. Misapplication of DCF techniques can certainly contribute to an unwarranted aversion to making long-term investments. However, the problem lies not in the technique but in its misuse. Money has a time value in every economy, and cash is the life-blood of every business.

To evaluate cash flows (costs or revenues) generated in different periods requires a procedure for making comparisons. For evaluating and ranking investment proposals, whether they have short or long lives, and involve capital equip¬ment, R&X), or marketing expenditures, we need techniques that recognize that cash flows occur at different times. Discounting provides a rational and conceptually sound procedure for making such eval¬uations. Unfortunately DCF techniques, like computers, can yield impressive-looking but misleading outputs when the inputs are flawed. Managers with biased assumptions may end up with biased conclusions.

The fault, however, lies not with the technique but with the analyst. The path to improved capital bud¬geting requires education in the proper use of rational techniques rather than their rejection out of hand. In our view, DCF techniques provide valuable in¬formation to assist management in making sound investment decisions. We emphasize the word assist Copyright © 1984 by the President and Fellows of Harvard College. All rights reserved. because it is people, rather than analytical tools, who make decisions. Managers may have many objectives and face many constraints in their decision making.

Nevertheless, they need information on the relative financial merits of different options. Properly em¬ployed, DCF techniques provide such information. The alternative is to ignore the time value of money and implicitly assume that, for example, a dollar earned ten years from now will have the same value as a dollar today. DCF procedures, as commonly applied, are subject to three serious pitfalls: Improper treatment of inflation effects, particularly in long-lived projects. Excessive risk adjustments, particularly when risk declines in later phases of a project.

Failure to acknowledge how management can reduce project risk by diversification and other responses to future events. Awareness of these pitfalls should help managers avoid uncritical use of DCF techniques that may lead to poor decisions. An R&D Project, for Example Although the comments here apply to a variety of investment proposals, we shall illustrate these three major pitfalls with the analysis of an R&D project. [Exhibit I lists examples of other investment projects that are frequently misevaluated for the reasons de¬scribed in this article. Because of their risk charac¬teristics, R&D projects present some especially thorny problems. The pronounced uncertainties in these projects affect the analysis of risk in many ways. As a result, R&D projects with acceptable— even exciting—risk/return profiles may fail to meet the payoff criteria that management has established. Let’s look at a typical (hypothetical) project that would be rejected on the basis of the incomplete DCF analysis common in industry today. Then we’ll show how a more complete and careful analysis reveals the project to be not only acceptable but highly desirable.

Our project has three distinct phases, as shown in Exhibit II. If the research (Phase 1) is successful, the project moves to market development (Phase 2), after which the resulting product may enjoy a long and profitable period of production and sales. The re¬search and market development phases are periods oi investment/ returns are forthcoming only during the third period (Phase 3) when the product is sold. Exhibit I Long-Term Risky Investments Frequently Misevaluated 1A consumer goods company considers test marketing the first of a proposed new family of products. A paper company studies investment in a new processing technique that could revolutionize paper making. 3A drug company looks at increasing its investment in biomedical research and the pilot plant that will be required if the research is successful. 4A real estate developer analyzes the first-stage investment in improvements at a greenfield site for industrial commercial facilities. 5A financial services firm considers investment in a telecommunications facility that could radically alter the future distribution of its services. A natural resources company evaluates a mineral-rights lease of a site that will require extensive development. It is important to differentiate between these phases, since each has decidedly different risk char¬acteristics. Market development (Phase 2) will not be undertaken unless the research (Phase 1) is suc¬cessful; thus, considerable uncertainty disappears be¬fore Phase 2 proceeds. Similarly, the sales period (Phase 3) follows only after successful results from research and market development. The information from Phase 2 will refine market projections, and Phase 3 cash flows are relatively low risk.

In sum, uncertainty about the project diminishes progres¬sively as we acquire more information. According to the probabilities shown in Exhibit II, the project viewed as a whole (rather than by phases) has the expected-value cash flows shown in Exhibit III and an expected internal rate of return (IRR) slightly over 10%. This appears distinctly unattrac¬tive, even ridiculous, when compared with custom¬ary rates of return (hurdle rates) of 20% or more for high-risk projects. Given this analysis and results, most managers would almost ertainly reject the project unless other strategic reasons dictated the investment. Many (if not most) U. S. companies, unfortunately, would probably analyze the project in this way, con¬cluding that it is indeed risky and has an expected IRR below normal hurdle rates. The interpretation of these “facts” is far from obvious, however, and requires a deeper understanding of DCF calculation procedures. The issue is not which buttons to push on a calculator, but rather the appropriate interpreta¬tion of the inputs and consequent output since the DCF procedure is no more than a processing tech-

HARVARD BUSINESS REVIEW January-February 1985 129 Exhibit IIProject Description Phase 1Research or product development $18 million annual research cost for 2 years 60% probability of success Phase 2Market development Undertaken only if product development succeeds $10 million annual expenditure for 2 years on the development of marketing and the establishment of marketing and distribution channels (net of any revenues earned in test marketing) Phase 3Sales Proceeds only if Phase 1 and Phase 2 verify opportunity Production is subcontracted

The results of Phase 2 (available at the end of year 4) identify the product’s market potential as shown below: Product demandProduct Annual net Probability life cash inflow High20 years $24 million . 3 Medium10 years $12 million . 5 LowAbandon None . 2 project Note: For simplicity, we assume that production is subcontracted in Phase 3 and that all cash flows are after tax and occur at year end. This assumption permits us to ignore some potentially complex tax issues involving epreciation and financing strategies. While a radical departure from reality, this assumption allows us to focus on issues of cash flow timing ana risk that appear to be less widely understood. Exhibit IIIExpected Cash Flows for the Project in $ millions YearsExpected value calculations 1- 18 2- 18 3. 6(- 10) = – 6 4. 6(- 10) = – 6 5-14. 6 (. 3 x 24 + . 5 x 12) = 7. 92 15-24. 6 (. 3 x 24) = 4. 32 Expected IRR = 10. 1% Adjusting for Inflation The most obvious shortcoming of the analysis is that it ignores how inflation will affect the various cash flows.

At one extreme, they may not be affected at all. On the other hand, the cash flows may adjust directly and completely with inflation, that is, an 8% inflation rate next year will raise cash flows in that and following years by 8 %. Most likely, inflation will affect different components of the cash flows in different ways and, when aggregated, the cash flows will adjust partially with inflation. Meaningful inter¬pretation of the calculated IRR requires knowledge of this inflation adjustment pattern. If complete adjustment were anticipated, the cal¬culated IRR would represent an expected real return.

However, comparing such real returns with nominal hurdle rates—including inflation—or nominal in¬vestment yields (for example, from government bonds) is not appropriate. 2 Historically, real yields on low-risk investments have averaged less than 5%, and the real yield on short-term U. S. Treasury securi¬ties has equalled close to zero. For higher risk invest¬ments, a frequent standard of comparison is the return (including dividends) on the Standard &. Poor’s “500” stock index. Over a 53-year period (1926-1978) the real rate of return on the S&P “500” averaged 8. %. While we cannot be certain that history will repeat itself, long-run averages do provide one stan¬dard for comparison. Since listed securities represent an alternative investment, projects of comparable risk reasonably should have expected returns at least as great. Returning to our hypothetical project, if cash flows adjust fully with inflation, the project offers a real return greater than the historic 8. 5% of the ScxP “500. ” Many types of cash flows, of course, do not adjust fully with inflation, and some do not adjust at all.

For example, depreciation tax shields, many lease payments, fixed-rate borrowing (like debentures), and multiyear fixed-price purchase or sales contracts do not change with the inflation rate. Consequently, a proper analysis requires an understanding of the inflation adjustment patterns for different cash flow segments. While American managers’ awareness of the im¬pact of inflation on project evaluation has risen in the last decade, even today many of them have at best a cursory understanding of it. Failure to incorporate inflation assumptions in DCF analyses can be par¬ticularly troublesome in decentralized companies. iique. The analysis appears sophisticated with its use of probabilities and discounting, but it is incom¬plete and seriously misleading. 130 2. lames C. Van Home, “A Note on Biases in Capital Budgeting Introduced by Inflation,” Journal of Financial and Quantitative Analysis, Janaury 1971, p. 653. HARVARD BUSINESS REVIEW January-February 1985 Corporate financial officers commonly specify com-panywide or divisional hurdle rates based on a cur¬rent (nominal) cost of capital. Furthermore, analysts at the plant or division level often estimate future cash flows (particularly cost savings) based on cur¬rent experience.

Unless those analysts consciously include anticipated inflation, they will underesti¬mate future cash flows and, unfortunately, many good projects may be rejected. Parenthetically, the converse is unlikely to occur: it is hard to conceive of an analyst using inflated cash flows with real discount or hurdle rates. Also, projects that go forward usually undergo several re¬views that are likely to result in some tempering, or lowering, of overly optimistic cash flow assump¬tions. By contrast, rejected projects are seldom given subsequent reviews that might reveal unrealistically low inflation assumptions.

The mismatch of inflation assumptions regarding cash flows and hurdle rates is generally most pro¬nounced for projects with payoffs years down the road. So long as the inflation rate is positive (even if declining), the gap between projected real cash flows and their nominal equivalents grows with time. For example, suppose that inflation rates for the next three years are expected to be 8, 6, and 4% respec¬tively. Consider an item that sells for $1 now. If its price will increase at the rate of inflation, its nominal price should be $1. 08 next year, $1(1. 08)(1. 6) = $1. 14 in two years, and $1(1. 08)( 1. 06)( 1. 04) = $1. 19 in three years. These inflated prices, rather than the current $ 1 price, should be incorporated into the DCF analysis if discounting is to occur at nominal rather than real interest rates. The error that arises from the failure to include inflation in cash flow estimates compounds with time as long as inflation is positive. Under these circumstances, distant cash flows, such as those characteristic of research and development invest¬ments, have present values that are more seriously understated.

It is difficult to know how widespread such errors have been during recent years, but almost surely they explain in part the shift toward shorter lived projects and myopic investment decisions in many businesses. ferent stages of the project, the investment becomes much more attractive. A typical discount rate (k) used in DCF analyses may be viewed as composed of three parts: a risk-free time value of money (RF), a premium for expected inflation (Eir), and a risk premium (A) that increases with project risk. This relationship can be repre¬sented as: 1 + k = (1 + RF)(1 + Eir)(l + A)

For example, a risk-free rate of 3% with 10% ex¬pected inflation and a 6% risk premium would imply 1 + k = (l. Q3)(1. 10)(1. 06) = 1. 20, or a nominal dis¬count rate of approximately 20%. Since inflation, as well as project risk and even the risk-free rate (RF), can vary over time, we should permit k to have different values at different times. The subscript t indicates the relevant time period; thus kt is a function of the RFt, Eirt; and At values for that period. To focus on situations where project risk is expected to change significantly through time, we will use real (deflated) cash flows and real dis¬count rates with RF constant.

It is, of course, very important to adjust for expected inflation properly. Without losing sight of that point, let’s shift the focus of discussion to risk adjustments by assuming that the inflation adjustments have been executed prop¬erly. Denoting the real (risky) discount rate for period t as rt, we have: 1 + rt = (1 + RF)(1 + At) This differs from kt simply by the removal of the inflation factor (1 + ETT,). Then by definition, the NPV of a project with expected real cash flows (CFt) occurring in two periods is: CF, 1 + r, (1 + r,)(l + r, CF, (1 + RF)(1 + A,) CF2 NPV = -^- + 1 + RF)2(1 + A,)|l + A2) + Avoiding Excessive Risk Adjustments A second flaw in the original DCF calculations for our hypothetical R&D project is the use of a single discount rate (IRR) for a project in which risk declines dramatically over time. As a result, the project ap¬pears less attractive than it really is. If we make appropriate adjustments for the differing risks in dif- CF, CF3 + This brings us to a key point. If A, = A2 = A, this formula collapses into the familiar form with a single discount rate: NPV = (1 + RF)(1 + A) (1 + RF)2(1 + A)2 CF, + CF2 1 + r (1 + r)2

In practice, virtually all DCF calculations are per¬formed using a constant discount rate such as r. In- HARVARD BUSINESS REVIEW January-February 1985 131 deed, financial calculators are programmed that way. Under what conditions, however, can we assume that A, = A2 (even approximately)? This assumption is reasonable if we anticipate that errors in predicting real cash flows result from a ran¬dom walk process—that is, predictions one period into the future always entail the same uncertainty. Thus if we were at time 1, each dollar of real cash flow in period 2 would look just as risky as each dollar of CF, looks now.

However, predicting two periods into the future is more risky,- thus CF2 viewed from the present deserves a larger risk adjustment. Consequently, CF2 is multiplied by 1/(1 + A)2 as opposed to simply 1/(1 + A) for CF,. In more general terms, the risk adjustment factor for a cash flow t period in the future is 1/(1 + A)’. The risk adjustment grows geometrically with time. Using a single risk-adjusted discount rate, there¬fore, implies an important and somewhat special as¬sumption about the risks associated with future cash flow estimates: such risks increase geometrically with chronological distance from the present.

On the infrequent occasions when this assumption is mentioned, it is usually justified on the grounds that the accuracy of our foresight decreases with time. While that argument has merit, consider what can happen when an investment proposal does not fit this pattern. Recall our R project. If the cash flows of Exhibit II are in real terms, the project has an expected real IRR of 10%; but there is a 40% chance of investing $36 million (real, after tax, but undiscounted) during the first two years and receiving nothing. Many deci¬sion makers would demand a much higher return than 10% (real or otherwise) to undertake such an investment.

If the project proceeds to Phase 3, the cash flows in that phase are considered relatively low risk. The large risk adjustments that were appro¬priate for early phases are no longer appropriate once we reach Phase 3. To highlight this point, let’s suppose that Phase 3 could be sold if the project successfully proceeds through the first two phases. Given its low risk, po¬tential investors might evaluate Phase 3 with a low discount rate such as 5% (real). Suppose market re¬search reveals a high demand for the product during Phase 3: 20-year life with annual net cash inflows of $24 million.

Discounting these flows at 5%, we reach a value at the beginning of Phase 3 (end of year 4) of $299 million. Thus if strong demand develops for the product, it’s possible the rights to produce and market it could be sold for a considerable sum. This value depends, however, on the marketing results from Phase 2. Thus we need to check what happens if less favorable demand conditions are revealed in Phase 2. Performing similar calculations for the other possible market conditions, we obtain the values in Exhibit IV. Even though there is a 20% chance of low demand, the overall expected value of selling Phase 3 is $136 million.

Suppose we now recalculate the project’s expected IRR assuming such an outright sale of Phase 3 for its expected value: $136 million. Using the 60% probability of Phase 1 success, we calculate the ex¬pected cash flows to be those in Exhibit V. Those net expected cash flows are equivalent to an expected IRR of approximately 28%. In other words, the pros¬pect that Phase 3 could be sold as just discussed leads us to revise the overall expected IRR for investing in the project from 10 up to 28 %. Since these calcula¬tions are in real terms, the project now appears quite attractive.

Pushing this analysis one step further, let’s assume the project could also be sold at the end of Phase 1 if the research is successful. That is, the new owner after purchasing the project would pay an estimated $10 million per year of Phase 2 costs and receive the Phase 3 value (depending on marketing research results) as shown in Exhibit IV. The purchaser would now encounter the expected cash flows indicated in Exhibit VI. Clearly this proposition is riskier than just buying Phase 3, since the marketing research results of Phase 2 are not yet known.

Suppose a potential pur¬chaser evaluated the cash flows in Exhibit VI using a 20% discount rate (well over twice the historic real return on the S “500”). The implied purchase price (present value at the beginning of Phase 2) is slightly over $79 million. But what is the implied return to the first owner—the initial developer of the product who undertakes the risky proposition of investing $18 million for each of two years in research—if a successful project could be sold at the end of two years for $79 million?

The expected real return (including the 40% chance of Phase 1 failure) is over 63%—a far cry from our initial estimate. This analysis illustrates a pitfall in evaluating proj¬ects with risk patterns that differ significantly from the simple random walk assumption. In our example, uncertainty is greatest during the first two years. But it is unreasonable to penalize more than 20 years of subsequent cash flows for that risk. To dramatize this point, we have assumed that the project can be sold in its latter phases.

Indeed, the project acquires a dramatically high value if Phase 1 succeeds—a point missed by the initial IRR calculation, which implic¬itly discounted all cash flows at the same rate. The difficulty with using a single risk-adjusted dis¬count rate (or IRR) is that the analysis blends time discount and risk adjustment factors. Unless project 132 HARVARD BUSINESS REVIEW January-February- 1985 i I Exhibit IVAnticipated Phase 3 Values if Sold in $ millions DemandProbabilityValue of Phase 3 year 4 High. 3299 Medium. 593 Low. 20 Expected value =136

Exhibit VExpected Cash Flows with Phase 3 Sale in $ millions YearOutflowInflowNet 1- 18- 18 2- 18- 18 3- 10 x . 6- 6 4- 10 x . 6136 x . 675. 6 Exhibit VIExpected Cash Flows for Purchaser of Phase 2 in $ millions YearOutflowInflowNet 3- 10- 10 4- 10136126 Exhibit VII Expected Cash Flows if the Project Can Be Abandoned During Phase 1 in $ millions YearExpected outflowExpected inflowExpected net cash flow 1- 18- 18 2- 78 x . 879 x . 633 risk follows a simple random walk pattern, this blending is inappropriate. Although this problem is discussed in the academic literature,3 it is enerally ignored in practice. For projects with dramatically different risk phases, the result can be a serious mis-estimation of project value. A more appropriate procedure for evaluating such projects is to separate timing and risk adjustments using the concept of certainty equivalent value (CEV). The CEV of a cash flow in a given year is simply its risk-adjusted value in that year. If we con¬verted all future cash flows to CEVs, we could then discount the CEVs to the present using a single risk-free discount rate.

With the timing and risk adjust¬ments thus separated, we avoid the possibility of compounding risk adjustments unintentionally. As a practical matter, attempting to convert each year’s cash flow into a CEV can be cumbersome since the CEV for period t may depend on probabilities for cash flows in the previous period (t-1), which in turn depend on probabilities from t-2, and so on. In -our example, the cash flows in Phase 3 depend on results from Phases 1 and 2. Indeed, we have assumed that management would abandon the project altogether if the research is unsuccessful or market tests indi¬cate low demand.

Although it is important to consider interactions among cash flows in different periods, the analysis of all possible management responses or other con¬tingencies would be extraordinarily complex and un¬wieldy. Thus we need reasonable approximations. Managers and analysts must exercise judgment re¬garding which risks and possible actions should be included in the analysis. We recommend that high-risk projects be evaluated as a sequence of distinct risk phases (of perhaps several years each). In our example, we did not attempt to calculate CEVs for each year in Phase 3.

Rather, we estimated a value for the whole phase conditional on the demand level. Similarly, our calculated $79 million value for the project if Phase 1 succeeded is a CEV (at the beginning of year three) for Phases 2 and 3 combined. In both cases, these CEVs are estimates of the proj¬ect’s potential selling price—its market value at the end of years two and four respectively. While the project might be worth more to the company if it retained all phases, the market CEVs represent op¬portunity costs for retaining Phases 2 and 3 that are useful (and conservative) yardsticks for evaluating the entire project.

Estimating market values for different phases is 3. See, for example, Alexander A. Robichek and Stewart C. Myers, “Conceptual Problems in the Use of Risk-Adjusted Discount Rates,” Journal of Finance, December 1966, p. 727. HARVARD BUSINESS REVIEW January-February 1985 133 I obviously an imprecise process. Using a single risk-adjusted rate for an entire phase (rather than separate rates or CEVs for each cash flow) produces only an approximation, unless risks within that phase have a random walk pattern.

The approximation is reason¬able, however, it the discount rate is low and/or the phase covers a fairly short period of time (as in Phases 1 and 2 above). If a phase is both long and risky, analysts should divide it into subphases. To restate our argument, we recommend seg¬menting projects into risk phases, then valuing se¬quentially each phase, working backward from the last. This procedure can be used to determine either an expected IRR on the initial phase (as already illus¬trated) or an NPV for the project. In general, we prefer calculating NPVs since this avoids technical prob¬lems with IRR, including scale ambiguities.

Al¬though slightly more complex than a standard expected NPV or IRR calculation, our approach is not difficult per se. It simply entails a short sequence of expected NPV calculations using different interest rates to value different risk phases. When a project’s risk pattern differs substantially from the simple ran¬dom walk assumption, such differences should be recognized and the evaluation procedure modified accordingly. As we have shown, evaluation based on inappropriate analysis can be very misleading. Considering the Eye of the Beholder

A third major problem in project evaluations is correctly assessing project risk and how managers can influence its nature and level. Here it is im¬portant to consider tho perspective of the analyst. Risk that seems excessive to an R&D or project man¬ager may appear reasonable to a corporate executive or a shareholder who can diversify the risk by spread¬ing it across other investments. Also, managers can influence the level of risk by future actions that affect the ultimate payoff of a project investment. Frequently, the major uncertainty in R&D invest¬ments is hether the research phase will produce a viable product. From the perspective of financial market theories such as the Capital Asset Pricing Model (CAPM), risks associated with the research phase are apt to be largely diversifiable. Conse¬quently, a public shareholder with a well-diversified securities portfolio will probably voice little or no concern about these risks. Success or failure in the lab is probably correlated weakly (if at all) with broad economic forces or other systematic nondiversifiable factors shjzt effect returns in the stock maiket as a whole. The CAPM and related theories stress that a proj- 34 ect’s total risk normally contains both diversifiable and nondiversifiable components. To the extent shareholders can easily diversify their holdings in the financial markets, they can reduce their portion of the project’s diversifiable risk to a very small level. Under these circumstances, the shareholders need worry only about the systematic portion of project risk. Thus a financial market approach suggests that the typical R&D project is much less risky from the perspective of a well-diversified public shareholder than it may appear to the individual performing the DCF analysis.

In contrast, managers, creditors, and even suppli¬ers may focus on total risk (including both diversifia¬ble and systematic components) at the company Calculating Inflation’s Effects To correctly allow for inflation in a DCF analysis, some analysts include it in the cash flows and use nominal discount rates. If inflation rates are expected to vary, different discount rates can be used for differ¬ent years in a net present value (NPV) calculation. Such a procedure, however, entails cumbersome cal¬culations.

Furthermore, consistency on a compa-nywide basis requires spcifying the annual series of discount and inflation rates to be used by analysts. The simple approach is to use a single “average” inflation rate with a single nominal discount rate, but this is not ideal. Although in many cases the distortion associated with this approximation is not serious, the pattern of cash flows and projected infla¬tion affects the size of the distortion. A preferable procedure is to use deflated cash flows with real discount rates. In this approach, analysts estimate the cash flow n each period, including the increase from inflation applicable to each of its seg¬ments (for example, zero for depreciation tax shields). Analysts then deflate the cash flow to present (for example, 1985) dollars using the projected inflation between now and that period. If the cash flow is expected to adjust fully with inflation, the deflation adjustment will exactly cancel the included infla¬tion. If not, the real value of that future cash flow will be altered by the extent to which it does not fully adjust with inflation.

The series of deflated (real) cash flows can then be discounted using real discount rates. Since the real time value of money appears to be considerably more stable than its nominal coun¬terpart, this second procedure is superior to using a single nominal discount rate. HARVARD BUSINESS REVIEW January-February 1985 I level. These groups have interests that are not easily diversified in the sense that the CAPM assumes. Thus they are concerned about total cash flow vari¬ability but at the company (not project) level.

Even at the company level, however, the R budget may be spread across many projects. A multi-industry company of even moderate size is probably suffi¬ciently diversified to allow large reductions in cash flow variability per dollar of R investment. Once again, the risk of a particular project appears lower from a portfolio perspective than from the perspec¬tive of an analyst looking only at the project itself. Most managers are aware of portfolio effects and the arguments regarding shareholder welfare based on financial market models such as the CAPM.

Nev¬ertheless, it is understandable that they view a proj¬ect with over a 50% chance of no payoff (as in our example) as highly risky. Under such circumstances, it is easy to ignore portfolio effects and worry too much about the risk of that particular investment opportunity. This excessive risk aversion is fre¬quently manifested in a too-high discount or hurdle rate, thus compounding the pitfalls already dis¬cussed. Analysts may also use conservative estimates: overestimates of development time or costs and un¬derestimates of both the magnitude and duration of subsequent payoffs.

Although the tendency toward excessive conservatism is both inevitable and diffi¬cult to overcome, management needs to be aware of its existence and sensitive to its consequences. As we said earlier, projects that have been rejected are seldom reevaluated. It is all too easy for a good project to be lost. While excessive risk adjustments are certainly not unique to R proposals, the problem may be more severe here because R projects involve large and obvious uncertainties. The key is that these risks are likely to be highly diversifiable.

Failure to recognize this fact represents a systematic bias against R projects. Managers can also affect the level of risk by influ¬encing the distribution of project payoffs. In our ex¬ample, there is a 30% chance that Phase 3 will be worth $299 million. There is not a symmetric chance of losing $299 million—because the company will abandon the project if faced with low product de¬mand. The result is an expected value for Phase 3 ($136 million) which is $43 million above the most likely estimate of $93 million.

Unfortunately, many project evaluations consider only the most likely cash flow estimates and ignore the asymmetry or skewness of the payoffs. This practice understates the project’s true value in situations in which, future management actions can improve profits or limit losses. This problem is more significant for R&D projects than for other investments because the company has greater flexibility to expand production for highly successful products and to abandon apparently un¬profitable efforts.

Such managerial actions can result in greater returns than estimated originally (larger revenues over a longer period) as well as reduced downside risk. In our example, suppose progress can be monitored throughout Phase 1, and management has the option to abandon the project at the end of the first year if certain goals are not met. If the probability of re¬search failure is equally divided between years one and two (20% each), the expected IRR from an initial investment in Period 1 research increases from 63% to 83%, with no change in our other assumptions [Exhibit VII shows the relevant cash flows).

Clearly management’s ability to skew a payoff distribution in the company’s favor can have an important influ¬ence on a project’s desirability. DCF Analysis in Perspective How much the misuse of DCF techniques has con¬tributed to the competitive troubles of American companies is a matter of conjecture. It is clear, though, that incomplete analysis can severely penal¬ize investments whose payoffs are both uncertain and far in the future. Given these perils, one might argue that DCF procedures should be avoided or should be accorded little weight in long-term invest¬ment decisions.

We strongly disagree. It is foolish to ignore or to indict useful analytical tools simply because they might be used incorrectly or incom¬pletely. Rather, analysts and decision makers should recognize potential problems and be careful to ensure that evaluations are performed correctly. Managers cannot treat a DCF evaluation like a black box, look¬ing only at the output. They need to break open the box, examine the assumptions inside, and determine how those assumptions affect the analysis of a proj¬ect’s long-term profitability.

DCF procedures can help evaluate the implications of altered price, cost, or timing assumptions, but man¬agers must first specify the correct assumptions. These procedures can also be used to examine the ef¬fects of different capacity expansion or R&D strate¬gies under many scenarios. However, again managers must specify the strategies or scenarios to be exam¬ined. In short, discounting is only one step in evaluat¬ing alternative investment opportunities. This fact has frequently been lost in the arguments (pro and con) about the use of discounting procedures. Blaming DCF procedures for shortsightedness, bi-