Benjamin C. LESTY is an asociste profesor at Harvard Business School in Boston Senne 1998 Improved Techniques for Valuing Large-Scale Projects BENJAMIN C. Esty ponsors typically use one of two methods to value equity investments in project finance deals. Either they value equity indirectly by discount- ing fee cash flows (FCF) using the weighted average cost of capital (WACC) and sub- tracting debt value; or they value equity directly by discounting equity cash flows (ECF) using the cost of equity (K,). In either case, most use a constant discount rate for cash flows in all periods. Despite their preva lence, these approaches represent simple tools that were designed for simple applica~ tions. Most project finance investments, however, are not simple valuation problems. (One feature that makes them complex is the fact that project leverage changes over time. For the typical project, the ratio of debt to total capitalization starts at 0%, rises to some- where in the neighborhood of 60%-85%, and then fills back down to 0% in later years. Because the cost of equity is a function of leverage, both it and the WACC will change as leverage changes. Thus the use of a single discount rate for all years is inappropriate. A second, and related, problem with the stan- dard approach is the measurement of lever- age. Even though valuation theory dictates the use of market value leverage, most peo- ple measure leverage using book values. Fail- lure to incorporate the effects of changing leverage or to measure leverage correctly can result in serious valuation errors. In the first section of this article, I describe these problems more fully and illustrate them using an equity cash flow valuation of a hypothetical project finance investment called PetroMexico. I then show how to address the problems using relative- ly simple techniques. The way to solve the problem of changing leverage is to use mul- tiple discount rates rather than a single dis- count rate; the way to solve the problem of ‘book value weights is to employ an alterna tive valuation technique called quasi-market valuation (QMV). It is worth noting that both solutions — multiple discount rates and quasi-market valuation — are aimed at improving dis- count rate estimation and have nothing to do ‘with improving cash flow projections. The failure to estimate cash flows accurately can also result in serious valuation errors. ‘Although I do not address cash flow estima- tion in this paper for the simple reason that cash flow estimation tends to be project spe- cific, I do show how Monte Carlo simula- tion can be used to analyze cash flow uncer- tainty. Continuing with my hypothetical example, I analyze PetroMexico’s net present value (NPV), debt service coverage ratio (DSCR), and most likely year of default as functions of oil price volatility. In the final section, I briefly discuss another new valuation tool called real options analysis, which can supplement dis- ‘counted cash flow (DCF) analysis for valuing large-scale projects. Unlike DCF analysis, “Tae Jounnat oF Project Finance 9real options analysis incorporates the benefits of man- agerial flexibility including the ability to defer action (not all investment decisions are “now or never” type decisions), to change strategy, and to abandon a bad project. Despite these benefits and the existence of well-developed theory showing that the real options approach is the right way to value large-scale invest- ‘ments, it can be difficult to implement accurately. As 2 result, it may take some time before it is generally accepted and commonly used in practice. Until that time, it is important to ensure that one uses the most accurate and efficient DCF tools possible, which is the point of this article. THE STANDARD APPROACH TO PROJECT VALUATION To illustrate the problems associated with the standard valuation approaches and their solutions, I present a hypothetical project called PetroMexico, a $2.02 billion oil-field development project in Mexi- co consisting of inland wells, a pipeline, and a coastal refinery. To build the facilities, the project requires capital expenditures of $300 million, $800 million, and $600 million over the first three years; these expenditures will be depreciated using straight-line depreciation over fifteen years (see Exhibit 1). The sponsors, two large integrated oil companies, will fund 64% of the project’ total cost with two rounds of non-recourse bank debt: $700 million in the first year and $600 million in the second year. The debt will carry a 10% coupon and will be fully amortized over fourteen years. The sponsors will fund the remaining 36% ($724 million) with equity invest- ‘ments in years 0, 1, and 2. Total investment during the first three years of $2.02 billion exceeds total capital expenditures of $1.70 billion by $324 million, This amount includes funds for start-up administra- tive expenses, interest expense during construction, net working capital, and a debt service reserve account (DSRA) equal to six months of principal and interest Beginning in year 3, the project will produce fifty million barrels of Maya crude per year at an oper- ating cost of $4.00 per barrel. One of the sponsors will buy all of PetroMexico'’s output at market prices assumed to be $11.40 per barrel (in year 3), Both the operating costs and revenues are assumed to grow at 1.0% per year. For simplicity, assume the project ends 10. iwwnoven Tecuques FOR VALUING LARGE-SCALE PROJECTS in year 25 when all of the oil is depleted; the tax rate is constant at 35%; and the costs and revenues are denom- inated in US. dollars Because the FCF/WACC and the ECF/K, approaches are theoretically equivalent, I could wre either method to value the equity invested in PetroMexico. I prefer the ECF/K, method for three reasons. First, it uses actual, estimated tax rates in the ‘eash flows rather than incorporating a single, constant tax rate in the discount rate, Projects that generate net operating losses, receive tax credits, or are located in countries with non-linear tax regimes will not have a constant, marginal tax rates over time. Second, the FCE/WACC approach assumes the net present value of debt is zero. Under this assumption, it is correct to assume that firm or project value minus debe value equals equity value (V ~ D = E). When this assumption does not hold, the FCF/WACC approach will result in errors. And finaly, it is conceptually much more diffi- cult to value multiple rounds of equity financing, par- ticularly if different sponsors invest at different times, using the FCF/WACC approach, Because it is easier to implement the ECF approach correctly, I focus on it throughout the rest of the article.? ‘The first step in ECF valuation is to calculate the equity cash flows assuming all residual cash flows are dis- tributed to the sponsors as dividends (see Exhibit 1). ECF = Cash available for debt service (CADS) — Principal payments = Interest payments — Equity investments where CADS = Earnings before interest and taxes (EBIT) + Depreciation = Cash Taxes — Capital Expenditures — Increases in net working capital (NWC) = Funds for the debt service reserve account (DSRA) ‘The next step is to estimate the expected cost of equity (K,) using the capital asset pricing model (CAPM), ‘According to the CAPM, the cost of equity, or the expected return on equity fom the investors’ perspective, is a fanction of the risk-free rate (R), an equity or levered beta (G,), and a market risk premium (Ry - Ry. Srnane 1999(vas) z96'8s 0 Tec'T9y —(er6'8v2) ior & Bee's 0 BILosh — (BL¥'9bz) ‘90FT % t6L'ts 0 zr'esh —(@e0'r¥2) zret @ wees 0 oss'crr — (z't¥2) eer w £59°9S (str'sst) 0 ver'erh soe w esos (aastest) 0 e9c'ecy iser oz Tes'ss — (Geo'zst) 0 see'ber eect 6 ales (BISTOST) «0 zso'0cr wel st Tepes scot) 0 r9e's8e el a o6s'es (esc'vor) 0 8671 oo esc'es 0 eL9'ty9 | SBzT st we'es 0 euz'9g9 ELT v1 tst'es 0 weeeo OTE et ULIS ° eus'ezo Lv a 17's 0 viv'Lt9 = SECT u 1scios 0 eases oectI9 = eat or 9ez'0S 0 609'6uz usz'so9 OFT 6 9pe'6r 0 699'scz swiss 86TT 8 1s¢'6r 0 908'tcz zor'ess =A t 198 0 996'192 eseuss = SCTE 9 9u'8r 0 19t's9z SIstss ev TE 5 96e'ty 0 soc'osz, esis «ISTHE + oze'ey 0 L99'9se, ooo'os «= OF'TIS € 6r8'9r (000'009) (eecred oo 0 z 0 (000'008) (00°02) 0 0 1 os os (o00'00¢8) os os os ° NODDY (MN) @xeL sompuedg (ayy seeuadeqg— anuaasy—(jouegued) 79, Baas rade yseyreudea, seep Superdog = OL. HOAPRID. 20189§ Suppom awoouy eke woaan »N ar07eq yop yo Surpuna sSuyureg payseoaiog (pave>ypur aspmzayjo ssajun ‘spursnowp Ut $57) Oorx=PVoxHed JO} WHOWIIES MOLT YseD PUE amMIRNS [ePUEULL I LIgtHxg “Tus Jounnat oF PRogscr France 1 src 1999ExHIBIT 1 (CONT’D) Financial Structure and Cash Flow Statement for PetroMexico (USS in thousands, unless otherwise indicated) Cash Debt Available Debt Service Service Equity for Debt Total Coverage ‘Cash Service Principal Principal Interest Debt Ratio Flow (CADS) Outstanding Payments _—_-Payments Service (scr ech ($300,000) 30 40 30 $0 ($300,000) (800,000) 700,000 0 70,000 (630,000) (170,000) (724,349) 1,300,000 o 130,000 (470,000) (254,349) 326,446 1,300,000 25,000 130,000 195,000 au 171,446 327,990 1,275,000 25,000 127,500 152,500 25 475,490 317,058 ‘1,250,000 25,000 125,000 150,000 2a 167,058 332,401 1,225,000 50,000 122,500 172,500 1.93 159,901 320,643 1,175,000 50,000 117,500 167,500 191 153,143, 335,161 1,125,000 75,000 142,500 187,500 179 147,661 322,523 975,000 75,000 97,500 172,500 1.87 150,023, 336,243 900,000 100,000 90,000 190,000 17 146,243 322,864 800,000 100,000 80,000 180,000 179 142,864 335,761 700,000 125,000 70,000 195,000 172 140,761 321,560 575,000 125,000, 57,500 182,500 176 139,060 333,636 450,000, 150,000 45,000 195,000 in 138,636 324,114 300,000, 150,000 30,000 180,000 1.80 waa 377,953 150,000 150,000 15,000 165,000 229 212.953 278,987 ° ° ° 0 278,987 281,798 o 0 0 ° 281,798 284,638 0 0 ° 0 284,638 287,506 0 ° ° 0 287,506 290,403 0 ° ° ° 290,403 293,330 ° ° 0 0 293,330 296,286 0 ° 0 ° 296,286 299,275 o ° 0 0 299,275 *DSCR = Cash Available for Debt Service (CADS) + Total Debt Service. Ret Be Ry - RD a) ‘The equity beta is, in turn, a function of the project's asset or unlevered beta (B,) and its leverage (V/E).? B, = BA(v/E) @ Here the leverage ratio is defined as firm value (V) divided by equity value (E), where firm value is the sum of debt and equity values (V = D + E). ‘The project’s asset beta can be estimated using 12, tummovan Tecntaques FOR VALUING LARGE-SCALE PROJECTS regression analysis and a set of comparable, publicly- traded firms. For the purposes of this analysis, I assume the asset beta for an integrated oil producer is 0.60. Given PetroMexico’s fanding mix of 64.2% debt (D/V) and 35.8% equity (E/V), its equity beta is: Be = B(V/E) = (0.60)(100.0%/35.8%) = 1.68 3) Plugging this value into the CAPM (Equation 1) yields an expected cost of equity of 20.4%, assuming a long- term risk-free rate of 8.0% and a historical market-risk Sonne 1999premium of 7.4% over long-term treasuries.? Kg = 8.0% + (1.68)(7.4%) = 20.4% “) Armed with a set of equity cash flow projec- tions and the cost of equity, itis staightforward to cal- ‘culate the project’ net present value. Assuming year 0 cash flows occur immediately (i.e. the discount factor — the reciprocal of the discount rate — is 1.000), the NPVis negative $60.3 million (see Exhibit 2). Because the NPV is negative, the sponsors should reject the project as failing to earn an appropriate risk-adjusted rate of return.* PROBLEMS WITH THE STANDARD APPROACH The standard valuation approaches were devel- ‘oped at a time when financial analysis was both time consuming and costly in terms of computing resources, and when firms had stable, long-run capital structures. To reduce the computational complexity, practitioners used a single, constant discount rate. This approach is incorrect because the project’s cost of equity is a func- tion of its equity beta which, in turn, is a function of leverage (sce Equations (1) and (2)). As seen in Exhibit 2, the project’s leverage begins at 0%, rises to 64% in year 2, and then falls back down to 0% in year 17. ‘When leverage changes, the discount rate must change. Exhibit 3A presents this phenomenon graphically. Cal- culating the project’s cost of equity using the point of ‘maximum leverage — typically the project’s finding mix — causes the cost of equity to be overstated for ‘most years because leverage is overstated for most years. Even simple adjustments, such as using the project's average debt ratio, are highly unlikely to yield the right answer (see Exhibit 3B).° This miss-estimation of the cost of equity also distorts the calculation of the WACC.* So whether you use the FCF/WACC or the ECF approach, you will get incorrect answers with a single discount rate. Never- theless, most valuation textbooks such as Ehrhardt [1994, p. 76], Finnerty (1996, chapter 7], and Copeland et al. [1996, pp. 249-250] recommend using the pro- Jject’s “target” capital seructure to calculate the appro- priate discount rate, a recommendation that will lead to errors when applied to project finance investments. The way to solve this problem is to calculate diferent discount rate for every year based on the lever~ Srmine 1999 age in existence at that time (see Exhibit 3C), an approach recommended by Damodaran (1994, p. 315) and Grinblatt and Titman (1998, pp. 323, 458]. For example, the correct discount rate for cash flows occur- ring in year two should be: Discount rate in year 2 = [1/(1 + Kp )JL1/(1 + Ky) 6) where Kg, and Kg, are the appropriate risk-adjusted costs of equity for years 1 and 2, respectively. Because the discount rate should incorporate cumulative risk through time, the following expression is not appropri- ate because it does not reflect the fact that year 1 risk may be different from year 2 risk (ie. if leverage differs over time): Discount rate in year 2 w+ KP 6) Exhibit 2 shows the valuation using the multiple discount rate approach. The cost of equity ranges from 12.4% when the project is fully equity funded to 20.4% when the project is at maximum leverage. The final col- umn shows the difference between the standard and multiple discount rate approaches: the cost of equity can, differ by as much as 8% per year! For people familiar with the FCF/WACC approach, the fact that increasing leverage causes the cost of equity to rise, but causes the WACC to fall may appear counterintuitive. This phe- nomenon happens for two reasons. First, as leverage and the cost of equity increase, the weighted average falls, because you are using less equity (the more expensive source of capital). Second, and more importantly, the standard WACC formula does not incorporate the effects of risky debt or the costs of financial distress, both, of which would cause it to increase with leverage. Interestingly, when the project is valued using the multiple discount rate approach, it has positive NPV of $11.4 million, which implies the sponsors should accept the project. This example illustrates how the standard approach can result not only in valuation errors, but also, in faulty managerial decision making. ‘What this alternative approach highlights is the critical links among leverage, equity risk, and equity returns, Because of this link, there is a need to calcu- late leverage ratios every year. According to finance theory, investors demand returns based on the market value of their investment, not on the historical book value. Except in very rare circumstances (when market ‘Tu Jounna oF Project France 13Srvc 1999 OTIS = AN (962098) = AaN 08 or200 za8'z 6000 vo | OO ore've, S66 me 08200 seve stioo voz | 00 rere, 989 os s1e00 sor orto evo | OO evel xo ssco0 Ig8' ssivo wre | 40 evevel 408 6st00 ors's zozoo voz | K0'0 ovevee %08 8400 169 vero voz | KOO etree «0's +0500 ove sooo wavoz, | OO overt 0's 19800 1586 eseoo roe | 00 erevel “08 19900 30's sewo wre | «00 ove'vee HO 41100 ween zisvo — seroe | eet orevel #9 z1800 Iss'8 tio YOR) MERE ove'vel 0s 82600 secor evo swwOe:SC| (ETRE ovetvee wry 690r0 isis 6800S aH] EH ereveL wet over cae’st eco. ware MTG ere'vel ae Lro0 0968 swore | «szs ove'vel sz 6s9r'0 oee'ee woe | rss Greve wz ¥oaco ost'ez savor | res HCW ast wero oee'ee woe | res orevel iad ozszo seu'ty voz | %9'09 eve'vee 90 99ce0 ouv'es sroe | 619 erc'vee x50 2z0r0 90099 sero. wwoz:S |Z erevee co caro sees sev. woz | MEED erc'vee 0 103s'0 sso roe | BCD ore'vee 400 94690 wav weroe | eH evel we oreo sow — ssron | BSS ooo'our 908 0000°r oot = sero | OO o0'00e8 en) Aiba (Ad eneA sone Gy Amba (Ad) mA JOP @p Gmbg wd) wd Aynba ety “Sita” mossia “Joey WaeaIg—wunooey 3010p OHIO TOL aoniog (@8ex0497 anfe A 100 (@Sera27 ane 00m) ane aouarayia —peoiddy soye rumored aidarowy weorddy prepuris oea (payeoapuy asyaurayyo ssaqun ‘spuesnowy ut eqT0 Sn) SBM anIeA Yoog Buss uORENIEA 4 Z LIGIHXY =values and book values are exactly the same in all time periods), one cannot use the capital structure mea- sured using book values. PetroMexico, like most approved investments, is expected to have a positive NPV, which implies that the market value of the pro- ject’s equity will exceed its book value. When this is tue, the overstated book value of leverage (D/V) will cause the cost of equity to be overstated and the resulting NPV to be understated. ‘The solution to this problem, at least in theory, is to measure the capital structure using market values. ‘Unfortunately, you do not know either the current or future market values of equity before you begin the project. In fact, the problem is inherently circular: you are discounting to get the present value of equity, but you need the present value of equity to calculate the right discount rate, A relatively new valuation tech- nique called quasi-market valuation (QMV) can be used to solve this problem. Richard Ruback of the Harvard Business School developed this technique to value highly-leveraged transactions such as leveraged buyouts (LBOs); yet QMV works equally well, if not better, for project finance investments. ‘There are three key assumptions underlying ‘quasi-market valuation, First, it assumes that the capi- tal asset pricing model works; in other words, that equity earns its expected rate of return in each period. As a result, the market value of equity at the end of period t (E,) equals the market value at the beginning of the period (E,,) plus the return on equity earned during the period (ROE, = E, Ky .,), adjusted for equity infusions (INV,) and dividend payments made during the period (D). E,=E,, + ROE + INV,-D, ” ‘The second assumption is that the book value of debt equals the market value of debt. This commonly- ‘made assumption holds except for firms in financial dis- tues. With this assumption, there is sufficient informa- tion to calculate a discount rate in all but the first year. Determining the capital structure at time zero requires the third assumption, an assumption regarding the effi- ciency of markets. The idea here is that market prices adjust to incorporate available information as soon as it is known. Accordingly, I assume that the projects NPV accrues to the initial equity holders on the day the pro- ject begins. Thus, the market value at the end of year 0 equals the initial equity investment plus the expected Sune 1999 EXHIBIT 3A Cost of Equity Calculated Using Maximum Leverage Leverage (debe) EXHIBIT 3B Cost of Equity Calculated Using Average Leverage ExHIBIT 3C Cost of Equity Calculated Using Changing Leverage ‘Tue Jounnat oF Propecr Founce 15NPV of the project which, of course, is not known at the time. Hence the circularity problem reappears. It can, however, be solved through iteration — for exam- ple, by using the “solver” function in Microsoft® Excel ‘The one thing that is known at time zero is that the project’s ending equity value in year 25 must be zero because the project ends.” Iteration will now solve the following problem: what initial net present value will ‘create a capital structure and concomitant discount rates such that the final equity value will be equal to zero and the discounted present value of equity cash flows will equal the initial present value of equity?® Exhibit 4 presents the quasi-market valuation of the PetroMexico project. Looking at year 1, the end- ing equity value of $627.280 million can be calculated as follows: Beginning Equity Value “+ Equity Investment + Return on Equity ~ Dividends = Ending Equity Value where 12.4% is the year 0 expected return on equity (technically 12.44% = 8.0% + 0.60 x 7.4% fiom the CAPM). Under the QMV approach, discount rates are based on the cost of equity at the beginning of the year — what equity holders expect to earn over the year — which is equal to the cost of equity at the end of the prior year. In contrast, most people use the end-of-year capital struc ‘ure to determine discount rates even though investors form expectations at the beginning of the year — the dif- ference can be significant when leverage is changing. ‘The computer runs through these calculations for all years, discounts the equity cash flows to get an NPY, plugs the calculated NPV back in as the project’ NPV in year 0, and repeats until the valuation con- verges on a solution. Each time the project NPV changes, the capital structure changes, the discount rates change, and the resulting NPV changes until con- version. In this case, the valuation converged to an NPV of $107 million, which is significantly above the NPV found using either the standard approach or the modified approach with multiple discount rates at book values. Based on the QMV approach, the managerial decision is to accept the project. ‘The final concern regarding the standard approach involves the use of discount rate adjustments 16 turnoven Teesisiques FOR VALUING LARGE-SCALE PROJECTS to reflect country-specific political risks such as expro- priation.? Academics such as Lessard [1996], and invest ‘ment bankers such as Abuaf and Chu [1994], Mariscal and Dutra [1996], and Godfiey and Espinosa [1997], recommend adding a country risk premium (CRP) to the cost of equity in Equation (1): K ot Bey -R) + CRP ® When available, most people use the spread between dollar-denominated sovereign debt (Yankee, Euro, or Brady Bonds) and US. Treasury bonds as a proxy for country-specific credit risk. For PetroMexico, the country risk premium would be the spread between ‘Mexican Brady bonds and ten-year US. Treasury bonds (Gee Exhibit 5). Ignoring the problems of trading liquidity, chang- ing spreads over time, and data availability (not all coun- tries have dollar-denominated sovereign debt) as imped ments to using this approach to measure country risk, there is the larger question of whether sovereign credit risk belongs in the discount rate in the first place. ‘According to valuation theory for integrated capital mar- kets, a project’ discount rate should incorporate only sys- tematic or market risk, while the cash flows should reflect tunsystematic or diversifiable risks. Technically the cash flows should be experied cash flows — i.e, the probabilis- tically weighted average of various future cash flow sce- natios.'® The key questions really are what risks are being considered and do they have a systematic component? In the case of expropriation, I do not know of evidence that shows it isa systematic risk, which implies that any adjustment for expropriation should be done to the cash flows.!! One can show, however, that adjusting the cash flows for a probability of default (ie., non-pay- ment for any reason including expropriation) is approx- imately equal to adding country risk premium to the discount rate (see the Appendix). Assuming A is the probability of default and CRP is the country risk pre- mium, then the approaches illustrated in Equations (9) and (10) yield approximately equal present values when the country risk premium is small and equal to the prob- ability of default (CRP = 2). Adjust the discount rate: T ECR o cea PVQ(ECR, (+K, +CRP} Sune 1999‘889°901$ = AaN 299%034 19s 61e00 PET ° (uz'eed) vier sarge sizosz es'or 6800 SKHZT vero (98e'962) 8zcz9 zec00s 98C 96H seer 90h PTT 222.005 gu'soz oxt'ese |e ser'er sho MEE 9su'S0L sre'sen sor 06s ievrt S00 «rz 8¥6'588 LEEW 905481 even HLS00 HTT 229 eH0"t GO'TRET BOVE Oe corer 99900 FZ coe 181't yecTOc'T B6L'Ise Gt zscoz 9200 PT vec't0e's ose'sop't Lg6'8cz 8 sect 91800 rzt zeeert es6zte Lt isch weet ZUTTeet HTT 9 avr wet zecese's 969'8CT St 9ss'9t weet spt'sec't 090%6ct Ht eut'6t wr zor'evel 19L0FF ot w9¢ wert ons'90z't t9e'@rt Zt sts St s99'291"T 96e'1€ 6st cor'ecr't zeciee eo i¥'v0r't ze'lr Lo £21690" 106'08 sre $39'000"F oct'e9 vst sa¢’6t0"t e189. et 618'S00 9a's6 ost $9200" #78601 rst £41066 (es'ze0) weet eros =O #601 Gre" pst o8'sz9 (or'ts) wre ows 0 Z6S'0S O0O'OLT 889904 (000'00¢8) wre sevoors 08 «== K=«=«OOTOOES SBNTHOTS OS waye opeg OO) RPA MOOR) GameuyseND) swuawxeE GOW wemseaul AAN (OE Sara wenoosi Aammba uo Apa THoL snrea Anbg pupa Ainby ymba welorg amtea nba Swed woes ania 3u wo Sujuuifog ust popada um Amba (paveorpuy asyauzaujo ssajun ‘spuesnoyy ul sreijop “S'/) WoENIEA woqrEUTFEND p LIGtHXg “Tu Journ or Paogecr France 17 Smane 1999EXHIBIT 5 ‘The Spread of Mexican Brady Bonds Over 10-Year U.S. Treasury Bonds seem true for large-scale projects. The probability of defaule is clearly not constant over time, (0317195) 02731793) capital structure is clearly not constant over time, and cash flows are not constant or perpetual. There are probably four reasons why this approach has ‘gained acceptance despite its the- oretical-flaws. First, like the-use ofa single discount rate discussed above, adjusting the discount rate greatly simplifies the analysis. It is far easier to adjust the discount rate by 2 constant term than to PP EP PELL LL PP LP EE OP ‘Year Adjust the cash flows T . Pv,(ECR)= £ ECRG= 2! (10) =I U+Kp) Use of Equation (9) (adjusting the discount rate) instead of Equation (10) (adjusting the cash flows) is essentially an admission to the fact that the cash flows are not expected cash flows, which violates basic valu- ation theory. In much the same fashion venture capital- ists add incremental risk premiums to counteract the effects of optimistic, “hockey stick” projections. The somewhat curious feature of this approach, however, is the belief that you can obtain corret values by using incorrect discount rates and incorret cash flows. Nevertheless, practitioners regularly use the adjustment shown in Equation (9) even though it is appropriate only when the following assumptions hold (Gee the appendix): 1. The probability of default or expropriation (A) is constant over time. 2. The project’s capital structure is constant over time, which implies that the expected cost of equity (K,) is also constant over time. 3. The project’ cash flows are constant and perpetual ‘Whaat is interesting about these assumptions is that none 18 mnoven Tecuiques FOR VALUING LARGE-SCALE PROJECTS generate multiple cash flow sce~ narios, determine relevant proba- bilities for each scenario, and cal- culate an expected cash flow sce- nario. Second, the projections ‘used for valuation purposes are usually derived from and used for operating budgets. The problem with “expect- ed” scenarios is that they do not resemble projects in any actual state of the world and it is time consuming to create multiple forecasts for multiple purposes. Third, Abuaf and Chu [1997] provide some evidence that local equity returns are related to local bond returns in ‘emerging markets. And finally, the errors introduced by changing the discount rate are typically not large com- pared to other potential sources of error (cash flow esti- mation or beta estimation).? And so, while adjusting the discount rate to reflect country risk is probably not the worst mistake one can make in valuing large-scale investments, itis probably better to adjust the cash flows under the theory that it is worth eliminating errors where possible. MONTE CARLO SIMULATION ‘One of the most important realizations for people responsible for valuation analysis is that the process is filled with error. As a result, sensitivity analysis isan inte- gral part of any valuation even though it is generally confined to static tests of particular inputs. A typical question for an oil-field project might be: “What hap- ens to the project’ NPV if oil prices are $8.00 instead of $11.40 per barrel?” Yet sensitivity analysis need not be confined to this kind of static analysis. ‘The alternative is to use Monte Carlo simulation smune 1999in which a computer generates a distribution of possi- ble outcomes based on a large number of scenarios Users can then analyze the fall distribution of possible ‘outcomes rather than a limited number of likely sce- ratios. For example, there might be two ways to struc~ ture a project with equivalent NPVs, but the NPV vari- ance or probability of default is lower under one struc~ ture. Managers are likely to prefer this structure over the other one. Tp illustrate the power of dynamic simulation analysis, I use a software package known as Crystal Ball® to analyze the effects of oil price volatility on PetroMexico’s financial risk and present value."® In par- ticular, I want to see how changing oil prices affect, PetroMexico’s NPV, debt service coverage ratio (SCR = CADS/Total Debt Service), and most likely year of default, ‘The first step is to pick a distribution for the change in Maya prices — one can think of price changes as returns on holding 2 barrel of oil. For this example, I assume oil prices follow a random walk. In particular, I assume oil prices have a lognormal distri- bution, which means that the continuously compound- ed return is normally distributed."* Given this distribu- tion, the next step is to generate random price changes, cover time: this year’s price (P,) equals last year’s price times one plus the random return: P, =P, (1 + random return) (ay ‘Aggregating annual price changes over time results in a random price path. The simulation software ‘uses these twenty-five-year price paths as inputs to the cash flow model in Exhibit 4 and calculates the various outputs of interest. Finally, the software repeats steps 2 and 3 1,000 times and generates a distribution of pos sible outcomes. For most applications, running 1000 scenarios takes only a few minutes, though more com- plicated scenarios such as quasi-market valuation can take longer. Exhibits 6A-6C provides the results from a Crys- tal Ball® simulation assuming the volatility of oil price changes is 20% per year and the project has 80% lever- age. Exhibit 6A shows the distribution of NPV’: the ‘mean is $472 million while the median is $108 million — note the median value is quite close to the quasi- market NPV of $107 million. The distribution clearly shows the option-like nature of leveraged equity: low oil Prices generate losses which are truncated by an amount Senne 1999 EXHIBIT 6A Project NPV Distribution Forecast Proeet Frequency Chart 1.000 Tats Mean =$472m vane Median = $108. Minimum ($298m) Be gE oy a ot tm oy EXHIBIT 6B Minimum Debt Service Coverage Ratio (DSCR) Distribution Forecasts Minimum DSCR 1000 Tate Frequency Chaet toutker ‘Cesetyie sr rm 09 say EXHIBIT 6C ‘Year of Default Distribution coy 012005 to 38 ey “Tue Journal oF Prowcr Fuunce 19EXHIBIT 7 Monte Carlo Simulation Results ‘Mean NPV $192m, Mean NPV $472m, ‘Median NPV 3110m, Median NPV = $108m Prob (NPV>0) 61% Prob (NPV>0) 55% 80% Mean Minimum DSCR = 113K, ‘Mean Minimum DSCR 0.94x Prob (DSCR (1) 78% Prob (DSCR (1) 57% Prob (Default) = 22% Prob (Default) 43% Leverage (debt/value) Mean NPV $214 ‘Mean NPV $470m, Median NPV $159m, ‘Median NPV $236m Prob (NPV>0) 67% Prob (NPV>0) 63% 50% Mean Minimum DSCR, 1.78 Mean Minimum DSCR 137% Prob (DSCR (1) 949% Prob (DSCR ( 1) 73% Prob (Defiuh) = 6% Prob (Default) = 27% Volatility of Oil Returns” “Standard deviation of annual changes in Maya oil prices. equal to the present value of the equity investment while high oil prices generate unlimited gains. Exhibit 6B shows the distribution of the minimum debt service coverage ratio (DSCR). Given the project’s high lever- age and oil-price volatility, it is not surprising the pro- Jject defaults 57% of the time (the “certainty” of having a DSCR (1.00X is 57.10% from the exhibit). The large number of observations at 1.00X illustrates the impor- tance of the debt service reserve account. If CADS is insufficient to cover total debt service, then the project relies on the debt service reserve account to cover the shortfall and the resulting coverage ratio is exactly 1.00X. The project defaults only when the CADS plus, the debt service reserve account cannot cover the total debt service. Finally, Exhibit 6C shows the distribution of default years in those instances where the project defaults. Given that the project has its lowest coverage ratios of approximately 1.70 in years 11, 13, and 15 (see Exhibit 1), it is interesting to note that the project is most likely to default well before that time. In fact, the 20. twmoven TecreugUes FoR VALUING LARGE-SCALE PROJECTS project is most likely to default in years 7 and 8. Having this kind of information allows project sponsors to restructure weak projects so as to minimize the probability of default and maximize expected equity returns:-For example, Exhibit-7 shows the impact of structuring the project with 50% leverage versus 80% leverage, under two oil price volatility scenarios (10% and 20%) Neither the mean NPV nor the probability that the NPV is greater than 0 are very sensitive to the leverage choice (though they are very sensitive to the volatility scenario — see Exhibit 7). However, the min- imum DSCR and the probability of default are both very sensitive to the amount of debt: in the 10% volatil- ity scenario, the minimum DSCR falls from 1.78X to 1.13X and the probability of default increases from 6% up to 22% as you increase leverage from 50% to 80%. As the probability of default increases, value declines as seen by the fact that the median NPV falls. These same effects appear in the high volatility scenario although their impact is magnified. Sain 1999EXHIBIT 8 Managerial Flexibility and Real Options Analysis & Managers: 5 wait and learn + Switch technologies Exhibit 7 alo shows how the project responds to different assumptions about oil price volatility. Intu- itively, the project should be worth mote, but default more often, as volatility increases. The reason NPV increases as oil price volatility increases is that the pay- ‘offs to levered equity resemble a call option: equity is protected on the downside by limited liability but enjoys unlimited upside. As seen in Exhibit 7, the mean NPV more than doubles in the high volatility scenar- ios. On the other hand, the mean minimum DSCR falls and the probability of default increases. At 50% leverage, the minimum DSCR fills from 1.78X to 1.37X while the probability of default increases from 6% to 27%, again consistent with intuition. Because Monte Carlo simulation shows the fall distribution of outcomes, it is far more informative than traditional, static sensitivity analysis. While I have shown the power of this analysis for one model input, namely oil price volatility, the model can easily be adapted to illustrate the effects of uncertain exchange rates, output quantities, construction costs, or input prices. Nichols [1994] provides a good description of how one company, Merck, uses Monte Carlo analysis as, part of its financial planning process. REAL OPTIONS ANALYSIS ‘The previous section illustrated the importance of understanding input uncertainty. What it did not do, however, is analyze how managers would respond to the Sram 199 ‘ Managers change course—> ‘Abandon (sp pan) various scenarios, which is a fandamental flaw inherent in all discounted cash flow analy- sis, In fact, authors such as Dixit and Pindyck [1995], Expand (mere of sm) ‘Trigeorgis [1996], Leslie and Michaels (1997), and Amram eee and Kulatilaka [1999] have Switch exhnobogies criticized DCF valuation because it makes several unre- CConeaet alistic assumptions. First, it assumes that investments are now-or-never decisions rather than recognizing that_man- agers have the option to wait. Often the key decision is not whether to invest, but when, Second, DCF analysis assumes that projects operate mecha- nistically; in other words, DCF assumes that managers are passive. Yet in reality, managers add value by making mid- ‘course corrections in response to new information — they have the option to change direction or strategy. Exhibit 8 illustrates the inherent flexibility managers have to defer investment and to change strategy in response to good or bad news having made an investment. Finally, DCF implicitly assumes that capital expenditures are reversible when, in fact, they are usually irreversible (.c., sunk costs). ‘The authors mentioned above advocate a new tool, called real options analysis, for valuing large-scale investments. While the term “options” refers to the fact that managers have flexibility when making invest- ment ot operating decisions, the term “real” is meant to distinguish these options on tangible and intangible assets from options on financial instruments. The advantage of real options analysis is that it explicitly recognizes and incorporates the value of being able to defer investment, expand output, change technologies, or stop investing. ‘Real option analysis is built on the recognition that there is a correspondence between managing real assets and investing in financial assets. For example, deciding whether to invest in PetroMexico is analogous, to deciding whether to exercise a call option on IBM. stock. If the sponsors spend the capital expenditures necessary to develop the field, then they will have the right to receive an uncertain stream of future cash flows from a developed oil field.!® They would only con- struct the field if the present value of expected future ‘Tae Journat oF Proyecr Finance 21cash flows exceeded the present value of development costs. Analogously, itis only rational to exercise a call option on IBM stock if the market price exceeds the exercise price, A related question, when to invest in PetroMexico, is akin to asking when it is rational to ‘exercise a call option, in this case an American call ‘option. The same tools that are used to value financial options, the Black-Scholes pricing formula, binomial trees, or numerical methods — can also be used to value teal options. While a full explanation of this new and exciting valuation tool is beyond the scope of this article, it is important to know that such tools ‘exist and are currently being used by companies like ‘Merck, Enron, and New England Electric to analyze major strategic investments (see Nichols [1994] and Corman (1997). Despite its theoretical appeal, the disadvantage of real options analysis is that it is significantly more complex to implement which explains its limited use to date. As academics and practitioners alike learn how to adapt this tool to the complexities of the real world, however, it will become more common. Until that time, it is important to make the best use of the tools available which, for better or worse, is the DCF approach. CONCLUSION ‘Valuing complex investments requires complex tools. Unfortunately most of the tools in practice today were not designed to handle the complexities of today’s investment decisions. In part, the problems lie in using the wrong tools — one should use real options analysis instead of DCF analysis — and in part the problems lie in using the existing tools incorrectly. The objective of this article has been to refine the project finance pro fessional’s valuation toolkit and to highlight some new tools. In particular, I discussed three ways to improve the discounting process in equity cash flow valuation (multiple discount rates, quasi-market valuation, and cash flow adjustments instead of ad hoc discount rate adjustments). T also discussed two new valuation tools, Monte Carlo simulation and real options analysis. ‘Admittedly, these tools and techniques are more diffi- cult to implement than the basic valuation tools people are used to, yet they provide more accurate valuations and, more importantly, better decision making, Given the low cost of personal computers and the relatively simplicity required to learn these tools, there is really 22 mwnoven Tecunaques FOR VALUING LARGE-SCALE PRORECTS little justification for not adopting them, APPENDIX Evaluating Country Risk — ‘The Country Risk Premium The present value (PV) of 2 seam of equity cash flows (ECF) equals: ECR, 1+ Kg) PY(ECH) (an) where the discount rate (K,) is the expected cost of equity 25 determined by the capital aset pricing model (CAPM). [fone adds 2 country risk premium (CRP) to the cost of equity, then the adjusted cost of equity (K,") becomes: Kip = Kp +ORP=R,+Be(R,-RY+ CRP (A-2) Substituting the adjusted cost of equity a a discount rate into Equation (A-1) yields PY(ECR = £ = a3) (14K, +CREY ‘Assuming the cath flows are constant over time (ECF = ECF, = ECF, ... ECF,), then Equation (A-3) is a stable perpetuity. ‘The present value when you adjust the discount rate is: ECE want Rate) = ECF _ PV(Ad} Discount Ratd = = (a4) Instead of adjusting the discount rate with country risk premium, one can adjust the cash flows to reflect a probabili- ty of default (A). Under the following assumptions, these two ‘methods are approximately equal mathematically: 1.. The probability of default (A) is constant over time, and Ossi; 2. The capital structure and, therefore, expected cost of ‘equity (KE) are constant over time, and 3. The cash flows are constant and perpetual, unless default coceuts in which cate they are zero. ‘With A at the probability of default, the expected equity cash flow in year t is: ECR, = ECF((t - A) + 0] 4 = ECF(I -2)" as‘which has a present value in year 0 of: Pvy(ECR) = ERG" as (+e) ‘Assuming constant cash flows beginning in year 1 in the absence of defiult (ECF = ECF, = ECF, ... ECF,), then, the cash flows with default form a geometric series: PV(AG} Cath Rows) = ECRIM) 4, BERIAA" | 4 7) O+K,) +k ‘which can be rewriten as PV(Ad} Cash Flows) = (1-2) ECF, ECRIAI | aK, +k (a8) Equation 8 is a growing perpetuity with 2 negative growth rate equal to A. It can be written as: PV(Adj, Cath Flows) = aw) ‘When the default risk is equal to the country risk pre~ mium (= CRE), then Equation (A-9) can be rewritten as: = pene 2) ain Substituting Equation 4 into Equation 10 yields: PV(Ad} Cath Flows) PV(Adj. Cath Flows) = (1 - CRP)PV(Ad). Discount Rate) ain Thus, the present value determined by adjusting the discount rate with a country risk premium (Equation (4-4) is approximately equal to the present value deter- mined by adjusting the cath lows for default risk (Equa- tion (A-10)) when the country risk premia (or probability of default) is small Smane 1999 ENDNOTES ‘The author thanks Mathew Mateo Millett for invalu- able assistance in writing this article and Rick Ruback for ‘many helpful discussions on valuing highly-leveraged transac~ tions including project finance transactions. The author also thanks the Division of Research at the Harvard Business School for supporting this research. "Both methods are prone to similar implementation errors. For example, most people use the promised or con- tractual debt rate, not the expected rate; in the WACC. Sim- ilarly, they derive equity cash ows by subtracting promised, not expected, debt payments from free cash lows. Both errors cause equity value to be understated. 2Equation (2) is the Hamada formula for levering and unlevering betas assuming rskless debt (ie., the debe has nei- ther interest-rate nor credit rsk which implies the beta of debt is zero, Bp = 0). Theoretically, the formula is: By = By X DMV) + BX (E/V), which reduces to Equation (2) when the debt beta is zero. For simplicity I have assumed rskless debt, 1 simplification that is easly corrected, Stbboton Associates (1998, Exhibit 2-1] reports that the difference in arithmetic means between market returns on. lage company stocks and long-term government bonds is, 7.4% for the period 1926-1997. ‘Note that I am calculating only the value of the pro- ject itself. To estimate the total value to the sponsors, I would need to include such things as the cost of providing debt guar- antees, if any, during construction as well as the interest tx. shields resulting fom additional sponsor borrowing against project cash flows. 5Using the average debt ratio of 32%, the cost of equi- ty is 14.6% and the NPV is $213 million. “The formula for the weighted average cost of capital (WACO) is: (D/V) x (1 -.) X Kp + (E/V)Kp, where (D/V) and (E/V) are the capital structure weights, Tc is the marginal corporate tax rate, Ky is the cost of debt, and Ke is the cost of equity. This assumption of zero ending equity value elimi nates the need to estimate a inal equity value which is why it is easier to apply QMV to project finance deals than to lever- aged buyouts (LBOs). "The QMV approach breaks down for projects with very negative NPVs because the negative NPV will exceed, the initial equity contribution causing the project to have 2 negative equity value which is not possible with limited lia- bility equity. Fanother common discount rate adjustment is the we of a “country beta” to adjust for emerging market volatility (1ee Lessard [1996] and Godfrey and Espinosa {1997)). Harvey [1995], and Bekaert and Harvey [1997] have shown that there is virally no relation between Thu Journal oF Progcr Fnance 23‘emerging market returns and betas measured with respect to a world market portfolio. Instead, emerging market returns are related to country volatility, which the authors hypothesize results from being tegmented from world cap- ital markets, Based on these results, certain practitioners advocate modifying the basic CAPM formula given in Equation 1 to reflect equity volatility Ke= Ry + an Be Ro) Where Buy = the emerging market beta = Pemwus ‘emerging market and the US. market, Ggy/Sus. = volatilities of the emerging market and the USS. market, respectively, Be the beta fora similar project in a devel- oped market like the US., and Rp = the market ris premium }0For example, if a project is expected to have cash flow of $100, but there is a 10% chance of expropriation, in ‘which case the cash flow will be $0, then the correct expect- ced cash flow is $90 = $100 X 90% + $0 x 10%. ‘\NDiamonte et al. [1996] provide some evidence of negative relation between political risk and stock returns, but do not show that political risk isa systematic risk "The errors are approximately equal to the size of the county risk premium (CRP), see the Appendix "Crystal Ball® (V. 3.0) isa graphically oriented fore~ casting and risk analysis software package which can be wed ‘with Microsoft Excel. Its produced by Decsioneering, Inc. (www decisioneering.com). ‘The continuously compounded return is defined as ner). ‘'SBrennan and Schwartz {1985} and Siegel, Smith, and Paddock [1987] have shown how option-pricing mod- cls can be used to value natural resource investments REFERENCES ‘Abuaf, Niso, and Quyen Chu. “The Executive's Guide to International Capital Budgeting: 1994 Update.” Salomon. Brothers Corporate Finance, August 1994. “The Intemational Cost of Capital — The Empirical Evidence.” Salomon Brothers Corporate Finance, June 1997. ‘Amram, Martha, and Nalin Kulatiaka. Real Options: Manag- ing Strategie Investment in an Uncertain Wold. Boston, MA: Harvard Busines Schoo! Press, 1999 Dekacrt, Geert, and Campbell R. Harvey. “Emerging Equi- 24 iurnoven Tecitwques FOR VALUING LARGE-SCALE PROFECTS ty Market Volatility.” Jouml of Financial Beonomics, 43 (1997), pp. 29-77. Brennan, Michael J, and Eduardo S. Schwartz, “Evaluating Natur Resource Investment.” Joumal of Business, Vol. 58, No. 2 (1985), pp: 135-157. Copeland, Tom, Tim Koller, and Jack Murrin. Valuation: ‘Measuring end Managing the Value of Companies, 2nd edition, John Wiley & Sons, Inc.: New York, 1996, Corman, Linda, “To Wait or Not To Wait." CFO Magazine, May 1997, pp. 91-94. Damodaran, Aswath. Damodaran on Valuation: Security Analy- sis for Investment and Corporate Finance. John Wiley & Sons, Inc.: New York, NY, 1994. Diamonte, Robin, John M. Liew, and Ross L. Stevens “Political Risk in Emerging and Developed Markets.” Finan- dal Analysts Journal, May/June 1996. Dixit, Avinash K., and Robert S. Pindyck. “The Options ‘Approach to Capital Investment.” Harvard Busines Review, May-June 1995, pp. 105-115. Ehchardt, Michael C. The Search for Value: Measuring the Com- pany's Cost of Capital, Harvard Business School Press: Boston, MA, 1994, Finnerty, John D. Project Financing: Asset-based Financial Engi- neering. John Wiley & Sons: New York, 1996. Godfrey, Stephen, and Ramon Espinosa, “A Practical ‘Approach to Calculating Costs of Equity for Investments in Emerging Markets.” Joumal of Applied Corporate Finance, 9 (3) (1996), pp. 80-89. Grinblatt, Mark, and Sheridan Titman. Financial Markets and (Corporate Strategy. lewin McGraw-Hill: Boston, MA, 1998. Harvey, Campbell R. “Predictable Risk and Retums in. Emerging Markets." Review of Financial Studies, 8 (3) (1995), pp. 773-816. Thbotson Associates, 1998 Yearbook, Stocks, Bonds, Bills, and Inflation, Leslie, Keith J,, and Max P. Michaels. “The Real Power of Real Options.” The McKinsey Quarterly, 1997, pp. 4-22. Lessard, Donald R. “Incorporating Country Risk in the Val- uation of Offshore Projects.” Joumal of Applied Corporate smune 1999)Finance, 9 (3) (1996), pp. 52-63. Mariscal, Jorge O., and Emanual Dutra. “The Valuation of Latin American Stocks: Part III.” Goldman Sachs Investment Research, November 26, 1996. Nichols, Nancy A. “Scientific Management at Merck: An. Interview with CFO Judy Lewent.” Harvard Business Review, Srnec 1999 January-February 1994, pp. 88-99. Siegel, Daniel, James Smith, and James Paddock. “Valuing (Offshore Oil Properties with Option Pricing Models.” Mid- land Corporate Finence Jounal, Spring 1987, pp. 108-116. Trigeorgis, Lenos. Real Options: Managerial Flexibility and Strat cay in Resource Allocation. MIT Press: Cambridge, MA, 1996. “Tw Jounnat oF Proyect Finance 25