Real Options: Valuing Waiting, Growth, Staged Investment, and Abandonment

A project forecast is incomplete if it assumes managers must follow the same plan regardless of what they learn. A real-options analysis values the ability to change that plan: wait, expand, stop, switch, or fund only the next stage. The essential questions are what can be decided, when, after learning what, and at what cost?

Learning goals and conventions

You should finish able to identify an actionable option, value a simple decision tree, distinguish flexibility value from total opportunity value, map a project into Black–Scholes under explicit assumptions, and evaluate a project-volatility estimate.

Use one currency, clearly dated cash flows, and costs net of the specified tax effects. Basic examples ignore financing complications. Where risk-neutral weights are supplied, assume a valid market-consistent pricing model; do not reinterpret those weights as subjective success probabilities.

Symbol	Meaning
$S$ or $V$	Current value of the underlying operating asset or opportunity
$X$	Investment cost to acquire or expand it
$T$	Time remaining to the decision deadline
$C$	Value of a call-like investment opportunity
$\sigma$	Annual volatility of underlying asset returns under the model
$r_f$	Risk-free rate, with compounding convention stated
$q$	Pricing probability in a one-step binomial model

1. A real option needs a decision, not just uncertainty

A real option is a right to make an operating or investment choice, rather than an obligation to follow a fixed plan. The underlying need not be a physical object: patents, licenses, platforms, research programs, and organizational capabilities can support real options too.

Two useful identification tests are:

1. Relevant information or changing conditions will become observable.
2. The firm can respond with a feasible action that changes cash flows.

A highly uncertain project is not automatically a valuable option. A sunk investment with no changeable operating choice may offer no relevant flexibility. A purported expansion option is not fully actionable if permits, competitors, contracts, or financing prevent expansion when needed.

Identify who controls the right and whether competitors share it. A nonexclusive opportunity can be eroded by entry. Maintaining the right can require spending, staffing, or renewal fees; these are part of its economics.

2. Common types of flexibility

Option	Decision after information arrives	Financial analogy and caution
Growth or follow-on	Launch the next product, sequel, or capacity stage	Call-like; the initial project must actually enable the follow-on
Defer	Invest later or not at all	American-call-like when exercise dates allow it; delay can sacrifice revenue or market position
Abandon	Close permanently and realize net salvage	Put-like; closure obligations and resale limits matter
Expand or contract	Change capacity or production scale	Incremental options with implementation costs
Temporary shutdown	Stop production and later restart	Different from irreversible abandonment; include restart costs
Staged investment	Pay for successive research or construction phases	A compound sequence of conditional choices
Switch	Change inputs, output mix, or process	Depends on relative prices, correlations, and switching costs

Options can interact. Abandoning a platform can destroy its expansion option; expanding can use capacity needed for another product. Adding stand-alone option values as if independent can overstate the value of the combined plan.

3. DCF is not the enemy: the fixed-policy approximation is

NPV remains the objective. The problem with a conventional fixed-plan forecast is that it may omit the ability to respond. A sufficiently specified state-contingent DCF or decision-tree valuation can include flexibility.

The other common error is discounting an option’s expected payoff at the underlying project’s unchanged WACC. An option has a nonlinear payoff, and its risk generally changes with the underlying value, time, and exercise policy. One static rate does not usually capture that exposure.

For a replicable call with stock-like underlying exposure and risk-free financing, a local approximation is

$$\beta_C\approx\frac{\Delta S}{C}\beta_S,$$

where $\Delta$ is the option’s exposure to a small underlying-price change. The elasticity $\Delta S/C$ can exceed one, giving the call greater systematic sensitivity than the underlying. It changes over time.

This does not mean every real option is riskier than its underlying or must have a positive high beta. An abandonment put can hedge downside; a bundle of operating assets and flexibility can have a different exposure again. Model the claim rather than applying a universal higher discount rate.

4. Worked option to wait: one-step replication

Suppose a traded or spanned operating asset is worth 100 today. In one year it will be worth 150 or 60 and pays no interim cash flow. A firm can pay 100 to acquire it either now or at the one-year deadline. The effective annual risk-free rate is 5%.

The pricing up probability is

$$q=\frac{1.05-0.60}{1.50-0.60}=0.5.$$

Check the asset price: $[0.5(150)+0.5(60)]/1.05=100$.

Decision at year 1	Strong state	Weak state
Underlying asset value	150	60
Pay 100 regardless of state	Net 50	Net -40
Invest only if worthwhile	Net 50	Net 0

Investing now has NPV $100-100=0$. The right to decide in a year has value

$$C_0=\frac{0.5(50)+0.5(0)}{1.05}\approx23.81.$$

Verify with a replicating portfolio

Hold

$$\Delta=\frac{50-0}{150-60}=\frac59$$

units of the underlying and borrow enough to owe $\Delta(60)=33.3333$ at year 1. Today’s cost is

$$\frac59(100)-\frac{33.3333\ldots}{1.05}\approx23.81.$$

In the strong state, assets worth 83.3333 minus debt 33.3333 leave 50. In the weak state, both are 33.3333 and the net payoff is zero.

The example assumes the right can be retained without a fee, competition, or lost interim revenue, and that pricing is supported by a traded spanning asset. If a license costs 5 today, net opportunity value falls to about 18.81. A positive option value is not the same as a free right available to every firm.

5. Staged investment: spend to learn, then choose

A pilot costs 10 today. At year 1, a follow-on operating asset will be worth 110 in a favorable state and 20 otherwise. Follow-on investment costs 60. Assume market-consistent state weights of 40% and 60%, a 5% risk-free rate, and no other cash flows.

After the pilot, management invests only in the favorable state:

$$V_{staged}=-10+\frac{0.4\max(110-60,0)+0.6\max(20-60,0)}{1.05} \approx9.05.$$

Committing to invest 60 regardless of the result would give

$$V_{committed}=-10+\frac{0.4(110-60)+0.6(20-60)}{1.05} \approx-13.81.$$

The incremental value of flexibility relative to that committed plan is about 22.86. The staged project’s net value is 9.05; do not add 22.86 to it again.

Staging can require extra setup costs, delay launch, or weaken scale economies. A pilot is useful when it changes later choices enough to justify its cost. Some uncertainty is resolved by waiting, while technical uncertainty may require an experiment; merely delaying does not always generate useful information.

6. Abandonment: compare continuation with net liquidation

At a future decision date, suppose continuing an existing asset is worth 130 in a good state and 40 in a bad state. Net liquidation value is 70. Under pricing weights one-half each and a 5% one-year risk-free rate, committed continuation is worth today

$$V_{continue}=\frac{0.5(130)+0.5(40)}{1.05}\approx80.95.$$

The ability to choose liquidation gives

$$V_{flexible}=\frac{0.5\max(130,70)+0.5\max(40,70)}{1.05} \approx95.24.$$

The abandonment option adds approximately 14.29. It pays $\max(70-V_1,0)$ relative to continuing. Do not add 70 of salvage to a continuation value that assumes the same assets remain in use.

If initial project cost is 90, committed NPV is approximately -9.05 while flexible NPV is +5.24. This does not imply every failing project should be liquidated: sale proceeds, shutdown obligations, restart possibilities, and lost growth options must all be included.

7. Static value, option value, and double counting

Define a baseline policy precisely. If the flexible strategy can always choose that policy and costs nothing extra to maintain, then

$$V_{flexible}\ge V_{baseline}.$$

A useful decomposition is

$$V_{flexible}=V_{baseline}+\underbrace{(V_{flexible}-V_{baseline})}_{\text{incremental flexibility value}}.$$

The second term must be measured relative to the same baseline. A Black–Scholes call value is often already the value of the whole optional follow-on opportunity, net of its exercise cost. Adding it to a DCF that already includes that same follow-on investment would double count.

For a base business with an expansion opportunity omitted from its forecast, adding the expansion-call value can be appropriate. For a forecast already containing optimal expansion decisions, an additional expansion premium is not justified.

A static DCF is a lower bound only when it is a feasible policy inside the flexible choice set, using consistent risk pricing, and without omitting extra flexibility-maintenance costs. An optimistic fixed forecast is not automatically a legitimate lower bound.

8. Mapping a project into Black–Scholes

For a simple option to acquire an operating asset at a fixed future date:

Project input	Option input	Check before using it
Value today of the operating asset to be acquired	$S$	Discount operating cash flows consistently; exclude the exercise cost
Investment outlay at exercise	$X$	Include required working capital and other incremental implementation costs
Decision deadline	$T$	Distinguish a contractual right from an opportunity competitors can remove
Matching risk-free rate	$r_f$	Use consistent maturity, currency, and compounding
Volatility of underlying asset returns	$\sigma$	Do not substitute revenue volatility, equity volatility, or a valuation standard deviation without a model

Black–Scholes assumes a particular diffusion, constant parameters, feasible replication, and other idealizations. A private project is not automatically a traded asset. Market and technical risks may not be fully spanned, so the mapping can be a decision aid rather than a uniquely determined arbitrage price.

Formula and worked example

For a European call on an underlying with no payout or value leakage before exercise, using continuously compounded $r_f$:

$$C=SN(d_1)-Xe^{-r_fT}N(d_2),$$ $$d_1=\frac{\ln(S/X)+(r_f+\sigma^2/2)T}{\sigma\sqrt T}, \qquad d_2=d_1-\sigma\sqrt T.$$

Let $S=100$, $X=110$, $T=2$, $r_f=5\%$ continuous, and $\sigma=25\%$. Then $d_1\approx0.1900$, $d_2\approx-0.1635$, and

$$C\approx14.23.$$

Immediate investment NPV would be $S-X=-10$, yet the optional opportunity can have positive value. This does not mean adding the call value to -10 is the correct total: the call already describes the optional investment opportunity.

A real timing option may allow earlier exercise. In the textbook no-payout traded-asset call, early exercise is not optimal under standard nonnegative-rate assumptions. Real projects can lose operating cash, patents, or competitive position while waiting, making earlier investment sensible. Those losses must be modeled rather than assumed away.

9. The two-number shortcut: interpret the dates correctly

In the same European no-payout model, define

$$A=\frac{S}{PV(X)}=\frac{Se^{r_fT}}{X},\qquad B=\sigma\sqrt T.$$

With annual effective risk-free rate $r_{eff}$, the equivalent $A$ is $S(1+r_{eff})^T/X$, where $r_f=\ln(1+r_{eff})$. Do not insert the same numeric rate into both compounding conventions.

For $B>0$:

$$d_1=\frac{\ln A+B^2/2}{B},\qquad d_2=d_1-B,$$ $$\frac CS=N(d_1)-\frac1A N(d_2).$$

A lookup table can therefore quote $C/S$ using $A$ and $B$. In the example, $A\approx1.0047$ and $B\approx0.3536$, giving $C/S\approx14.2340\%$.

$A$ compares operating value with a cost paid at date $T$. It is the PI for an otherwise committed future-date investment under those assumptions. For investment paid today, PI is $S/X$, and immediate NPV is $S-X$. Here $S/X\approx0.9091$, whereas $A>1$ because the future cost has PV about 99.53.

$B$ measures cumulative log-return volatility under the model’s constant-volatility diffusion. Square-root-of-time scaling is a modeling property, not a universal law for every business uncertainty.

10. Estimating project volatility

Industry judgment and sensitivity ranges

An informed range can indicate whether flexibility might be material. A classroom statement that 20%–30% is plausible for some projects is not a universal empirical estimate. Use the asset’s actual drivers, horizon, and definition; report sensitivity rather than selecting a high number to justify investment.

Comparable assets and traded options

Historical or implied equity volatility is volatility of a levered residual claim, not automatically the volatility of operating assets. Leverage can raise equity volatility, diversification across businesses can lower it, and cash holdings change the mix.

Consequently, a single project is not automatically more volatile than a quoted company’s equity. Adjust business mix, financing, cash, and horizon coherently. Also distinguish total volatility used in the option model from systematic beta used in a discount-rate model.

Simulation from operating drivers

Build a model for commodity prices, demand, costs, exchange rates, and their correlations. Translate them into operating cash flows and values at future decision dates. Infer a return process consistent with the intended option model.

A standard deviation of dollar NPVs is not an annual return volatility. Simulated forecast uncertainty is not automatically the same as tradable market risk, and parameter uncertainty need not behave like a lognormal diffusion. Risk-neutral simulation requires appropriate pricing dynamics; real-world simulation with subjective probabilities does not become risk-neutral by discounting at the risk-free rate.

Plain simulation values a fixed policy. Exercise choices generally require backward induction, a suitable regression method, or another optimization procedure. Include correlations and interactions instead of adding independently simulated option values.

11. Choose a method appropriate to the decision

Situation	Useful approach	Main limitation to examine
One or a few decision dates and states	Decision tree or binomial model	State probabilities and risk adjustment must be justified
Simple spanned investment call	Black–Scholes-style approximation	Payouts, changing costs, exclusivity, and replication assumptions
Several stages or switching choices	Backward induction or numerical optimization	Interactions and execution constraints
Unspanned technical risk	Scenario analysis with explicit preferences/pricing assumptions	No unique no-arbitrage price follows automatically

At an exercise date, compare all feasible actions, including retaining the option. In a spanned binomial model, a simplified recursion is

$$V_t=\max\left\{V_t^{exercise},\frac{qV_{t+1}^{up}+(1-q)V_{t+1}^{down}}{1+r_f} -\text{current waiting cost}\right\}.$$

All terms must be in date-$t$ value, with the cost timing stated. More general problems need additional actions and state variables. A managerial intention to wait has no value if the firm will lack authority, cash, or operating capacity to exercise later.

12. Practical interpretation

Higher volatility raises a textbook call’s value while other model inputs remain fixed. It does not imply that deliberately making an entire business more volatile creates value: uncertainty can damage base operations, funding capacity, counterparties, or the ability to exercise.

Similarly, longer expiry is useful only if the right survives and retaining it is feasible. Competition, loss of first-mover advantage, expiration, maintenance, and foregone cash flows can make delay costly.

A credible real-option proposal states the right, the information that changes the action, the exercise rule, the cost of preserving flexibility, and the evidence behind its valuation. “Strategic value” without those elements is not a substitute for a forecast.

13. Practice questions

1. A proposed factory is uncertain, but all spending is irrevocably committed before any information arrives. Does uncertainty alone establish an option to defer?
2. An underlying worth 80 moves to 120 or 50 in one year, pays no interim cash, and the risk-free rate is 5%. Find the binomial pricing probability. A right to invest 90 at that date pays what in each state, and what is its value?
3. A pilot costs 6 today. At year 1 the follow-on asset is worth 100 or 30, expansion costs 50, pricing weights are 0.3 and 0.7, and the risk-free rate is 5%. Find staged net value.
4. Continuing an asset at a decision date is worth 100 or 20. Net liquidation gives 40. What is the abandonment-option payoff in each state relative to continuing?
5. A base business is worth 200, excluding an expansion opportunity valued as a call at 15. Find combined value. What if the 200 already includes optimal expansion?
6. A project asset has $S=90$ and future investment cost $X=100$ in two years. With annual effective rate 5%, find $A$. Find immediate-investment PI. Are they the same timing comparison?
7. With annual volatility 30% and four years to exercise, find $B$. Why is a simulated dollar NPV standard deviation not directly interchangeable with 30%?
8. Explain why applying the existing operating WACC to all option payoffs can be wrong, and why the inability to trade or span a project matters for risk-neutral valuation.

14. Answers and reasoning

1. No. The claimed deferral flexibility must actually exist. Other operating options may exist, but they require a separately identified decision.
2. Up multiplier is 1.5 and down multiplier 0.625. $q=(1.05-0.625)/(1.5-0.625)\approx0.4857$. Payoffs are 30 and zero, so value is $q(30)/1.05\approx13.88$, conditional on the spanning assumptions.
3. Follow-on payoffs are 50 and zero. Staged net value is $-6+0.3(50)/1.05\approx8.29$.
4. Payoffs are $\max(40-100,0)=0$ and $\max(40-20,0)=20$. Flexible asset values are 100 and 40; salvage is not added to continuation in both states.
5. If excluded, combined value is 215 before any additional costs of obtaining the right. If optimal expansion is already included, adding 15 again double counts it.
6. $A=90(1.05)^2/100=0.99225$. Immediate PI is $90/100=0.9$. The first compares with a future cost’s PV; the second compares with a cost paid now.
7. $B=0.30\sqrt4=0.60$. Dollar valuation dispersion has different units and may reflect different horizons, discounting, or parameter uncertainty rather than the modeled annual asset-return process.
8. An option’s nonlinear exposure and leverage change by state and time, so its required return need not equal the underlying WACC. Without spanning, no-arbitrage alone may not determine a unique price; technical and other unhedgeable risks require additional assumptions rather than arbitrary risk-free discounting.

Further reading

• Dirk Jenter, MIT OpenCourseWare, Real Options, Lecture 19 Section A: identifying flexibility, project mapping, and the Oz Toys example.
• Katharina Lewellen, MIT OpenCourseWare, Real Options, Lecture 19 Section B: valuation assumptions, volatility estimation, and shortcut tables.
• MIT OpenCourseWare, Finance Theory II lecture notes: the progression from FCF and financing adjustments to real options.

Feynman study lab

Explain it simply

The ability to wait, expand or stop can change a project’s value.

Concrete example. A pilot costing 1 can preserve the choice to invest another 20 only if demand looks promising.

Teach it back without looking

Close the lesson and explain the idea aloud to a beginner in two minutes. Define each technical term you use, walk through the example, and state one assumption or limitation. Then answer:

Why is the pilot not the same as committing 21 today?

Check your explanation after you try

The later 20 is conditional on new information. Flexibility has value only after accounting for pilot cost, delay costs and competitive effects.

Find the gap and try again

If you hesitated, relied on a formula you could not explain, or missed the self-check, write down the exact gap. Revisit the relevant section of this lesson and work through its example. Close the notes, explain it again in simpler language, then change one number or assumption and predict what changes. Recalculate or check the lesson to test your prediction. Try once more tomorrow from memory.