Incorporating Managerial Cash-Flow Estimates and Risk Aversion to Value Real Options Projects. The Fields Institute for Mathematical Sciences

Transcription

1 Incorporating Managerial Cash-Flow Estimates and Risk Aversion to Value Real Options Projects The Fields Institute for Mathematical Sciences Sebastian Jaimungal Yuri Lawryshyn University of Toronto, Toronto, Canada November 26, / 53

2 Agenda Introduction Motivation Real Options Matching cash-flows General approach (numerical solution) Normal distribution (analytical solution) Indifference pricing General approach (numerical solution) Normal distribution (analytical solution) Results Practical implementation Conclusions 2 / 53

3 Motivation To develop a theoretically consistent real options approach to value R&D type projects Theoretical Approaches: Cash-flow determined by GBM df t = µf t dt + σf t dw t Practice: Managerial supplied cash-flow estimates consist of low, medium and high values 3 / 53

4 Valuation of R&D Projects: Managerial Sales and Cost Estimates Managers provide sales and cost estimates Table : Managerial Supplied Cash-Flow (Millions $) Sales COGS GM SG&A EBITDA CAPEX Cash-Flow / 53

5 Standard NPV Approach Using CAPM Ryan and Ryan (2002) report that 83% of businesses apply the WACC to value discounted cash-flows (DCF) CAPM: E[r E ] = r f + β C (E[r M ] r f ) Use of CAPM implies beta: β C = ρ M,C σ C σ M 5 / 53

6 Standard NPV Approach Using CAPM Ryan and Ryan (2002) report that 83% of businesses apply the WACC to value discounted cash-flows (DCF) CAPM: E[r E ] = r f + β C (E[r M ] r f ) Use of CAPM implies beta: β C = ρ M,C σ C σ M Some assumptions regarding β when using WACC Market volatility, σ M, is known Cash-flow volatility: σ project = σ C ? Correlation of the cash-flows: ρ project = ρ C ? 6 / 53

7 Standard NPV Approach Using CAPM Ryan and Ryan (2002) report that 83% of businesses apply the WACC to value discounted cash-flows (DCF) CAPM: E[r E ] = r f + β C (E[r M ] r f ) Use of CAPM implies beta: β C = ρ M,C σ C σ M Some assumptions regarding β when using WACC Market volatility, σ M, is known Cash-flow volatility: σ project = σ C ? Correlation of the cash-flows: ρ project = ρ C ? Some further assumptions regarding DCF: No managerial flexibility / optionality imbedded in the project Financial risk profile of the value of the cash-flows matches that of the average project of the company 7 / 53

8 Standard NPV Approach Using CAPM Ryan and Ryan (2002) report that 83% of businesses apply the WACC to value discounted cash-flows (DCF) CAPM: E[r E ] = r f + β C (E[r M ] r f ) Use of CAPM implies beta: β C = ρ M,C σ C σ M Some assumptions regarding β when using WACC Market volatility, σ M, is known Cash-flow volatility: σ project = σ C ? Correlation of the cash-flows: ρ project = ρ C ? Some further assumptions regarding DCF: No managerial flexibility / optionality imbedded in the project Financial risk profile of the value of the cash-flows matches that of the average project of the company Proper beta: β project = ρ M,projectσ project σ M 8 / 53

9 Standard NPV Approach Using CAPM Ryan and Ryan (2002) report that 83% of businesses apply the WACC to value discounted cash-flows (DCF) CAPM: E[r E ] = r f + β C (E[r M ] r f ) Use of CAPM implies beta: β C = ρ M,C σ C σ M Some assumptions regarding β when using WACC Market volatility, σ M, is known Cash-flow volatility: σ project = σ C ? Correlation of the cash-flows: ρ project = ρ C ? Some further assumptions regarding DCF: No managerial flexibility / optionality imbedded in the project Financial risk profile of the value of the cash-flows matches that of the average project of the company Proper beta: β project = ρ M,projectσ project σ M Matching method uses managerial supplied cash-flow estimates to determine σ project 9 / 53

10 Real Options Why real options? Superior to discounted cash flow (DCF) analysis for capital budgeting / project valuation Accounts for the inherent value of managerial flexibility Adoption rate 12% in industry (Block (2007)) What is required? Consistency with financial theory Intuitively appealing Practical to implement 10 / 53

11 Introduction: Real Options Approaches * As classified by Borison (2005) 11 / 53

12 Relevant Literature - Utility Based Models Berk et al. 1 developed a real options framework for valuing early stage R&D projects Accounts for: technical uncertainty, cash-flow uncertainty, obsolescence, cost uncertainty Value of the project is a function of a GBM process representing the cash-flows Main issue: how to fit real managerial cash-flow estimates to a GBM process Miao and Wang 2, and Henderson 3 Present incomplete market real options models that show standard real options, which assume complete markets, can lead to contradictory results 1 See Berk, Green, and Naik (2004). 2 See Miao and Wang (2007). 3 See Henderson (2007). 12 / 53

13 Matching Method Advantages The approach utilizes managerial cash-flow estimates The approach is theoretically consistent Provides a mechanism to account for systematic versus idiosyncratic risk Provides a mechanism to properly correlate cash-flows from period to period The approach requires little subjectivity with respect to parameter estimation The approach provides a missing link between practical estimation and theoretical frameworks 13 / 53

14 RO in R&D Applications: Managerial Cash-Flow Estimates Managers provide cash flow estimates 14 / 53

15 RO in R&D Type Applications: Two Approaches Managers supply low, medium and high sales and cost estimates (numerical solution) Managers supply ± sales and cost estimates from which a standard deviation can be determined for a normal distribution (analytical solution) 15 / 53

16 RO in R&D Type Applications: Low, Medium and High Sales and Cost Estimates Managers supply revenue and GM% estimates Scenario End of Year Sales (Margin%) Optimistic (50%) (60%) (65%) (60%) (60%) (55%) (55%) Most Likely (30%) (40%) (40%) (40%) (35%) (35%) (35%) Pessimistic (20%) (20%) (20%) (20%) (15%) (10%) (10%) SG&A* 10% 5% 5% 5% 5% 5% 5% Fixed Costs * Sales / General and Administrative Costs 16 / 53

17 RO in R&D Type Applications: ± Sales and Cost Estimates End of Year Sales (Margin) Sales 52 ± ± ± ± 15 COGS (31 ± 6) (37 ± 7) (44 ± 9) (46 ± 10) SG&A 10% 5% 5% 5% CAPEX (30 ± 6) (25 ± 5) (20 ± 4) (20 ± 14) σ CF = σ S (Sales) σ C (COGS) σ EX (CAPEX) σ CF (Cash-Flow) σ 2 S + σ2 C + σ2 EX 2ρ S,C σ S σ C 2ρ S,EX σ S σ EX + 2ρ S,C ρ S,EX σ C σ EX 17 / 53

18 Real Options in R&D Type Applications Problem: How should we value the cash flows? How should we account for managerial risk aversion? Approach: Apply matching method with MMM to value cash flows Apply indifference pricing to determine the value with manager s risk aversion Why Account for Risk Aversion: MMM assumes investors are fully diversified Impact of managerial risk aversion on the valuation of a real options project can enhance decision making 18 / 53

19 Market Stochastic Driver Traded index / asset di t = µi t dt + σi t dw t Assume there exists a Market Stochastic Driver / Indicator correlated to the traded index ds t = νs t dt + ηs t (ρdw t + 1 ρ 2 dw t ) Market stochastic driver does not need to be traded could represent market size / revenues is not constrained to a GBM process Risk-neutral MMM di t = ri t dt + σi t d W t ( ds t = νs t dt + ρηs t d W t + ) 1 ρ 2 dw ν = ν ρη σ (µ r) 19 / 53 t

20 Match Cash Flow Payoff 20 / 53

21 Match Cash Flow Payoff Each cash flow is effectively an option on the market stochastic driver, V T = ϕ(s T ), and so, we match probabilities P(ϕ(S T ) < v) = F (v) P(S T < ϕ 1 (v)) = F (ϕ(s)) η2 (ν P(S 0 e 2 )T +η T Z < S) = F (ϕ(s)), Z N(0, 1) P P (Z < ln S S 0 (ν η2 2 )T ) η = F (ϕ(s)) T ( ln S S Φ 0 (ν η2 2 )T ) η = F (ϕ(s)) T 21 / 53

22 Match Cash Flow Payoff ϕ(s) = F 1 {Φ ( ln S S 0 (ν η2 2 )T η T )} 22 / 53

23 Information Distortion 23 / 53

24 Risk-Neutral Measure Theorem The GBM Risk-Neutral Distribution. The conditional distribution function F vk S t (v) of v k conditional on S t at t, for 0 < t < T k, under the measure Q is given by ( ) T F vk S t (v) = Φ k T k t Φ 1 (Fk (v)) λ k (t, S t ) where the pseudo-market-price-of-risk λ k (t, S) = 1 η T k t ln S + ν 1 2 η2 Tk t ν 1 2 η2 T k S 0 η η Tk t. Note that as t 0 and S S 0 then λ k (t, S) ρ µ r σ Tk, i.e. the valuation is independent of ν and η. 24 / 53

25 Option Pricing Value of the cash flows n V t = e r(t i t) E Q [V ti F t ] = i=1 n e r(t i t) E Q [ϕ i (S ti ) F t ] i=1 Value of the project with option V = e rt E Q [max (V t K, 0)] ( n ( = e rt K i=1 e r(t i t K ) S ti = S 0 e ( ν 1 2 η2 )t i +η( t K x+ t i t K y) ϕ i (S ti ) e y 2 ) ) 2 dy K 2π + e x2 2 2π dx 25 / 53

26 Matching Cash-Flows for Normally Distributed Estimates Assume that the managers have provided cash-flow estimates of the form N(µ k, σ 2 k ) Assume the Market Stochastic Driver to be a Brownian motion Assume that there exists a cash-flow process: F t Introduce a collection of functions ϕ k (S t ) such that at each T k, F Tk = ϕ k (S Tk ) Theorem The Replicating Cash-Flow Payoff. The cash-flow payoff function ϕ k (s) which produces the managerial specified ( distribution Φ s µk for the cash-flows at time T k, when the underlying driving uncertainty S t is a BM, and S 0 = 0, is given by σ k ) ϕ k (s) = σ k Tk s + µ F k. 26 / 53

27 Value of the Cash-Flows for Normally Distributed Estimates Theorem Value of the Cash-Flows. For a given set of cash-flow estimates, normally distributed with mean µ k and standard deviation σ k, given at times T k, where k = 1, 2,..., n, the value of these cash-flows at time t < T 1 is given by V t (S t ) = n k=1 and for the case where t = 0, ( ) e r(t σk k t) (S t + ν(t k t)) + µ k, Tk V 0 = n k=1 e rt k ( νσ k Tk + µ k ). 27 / 53

28 Option Pricing for Normally Distributed Estimates Theorem Real Option Value of Risky Cash-Flows Estimates. For a given set of cash-flow estimates, normally distributed with mean µ k and standard deviation σ k, given at times T k, where k = 1, 2,..., n, the value of the option at time t < T 0 to invest the amount K at time T 0 < T k to receive these cash flows is given by [ ( ) ( )] RO t(s t) = e r(t 0 t) ξ1(s t) K ξ1(s t) K (ξ 1(S t) K) Φ + ξ 2 φ where Φ( ) and φ( ) are the standard normal distribution and density functions, respectively, and n ( ) ξ 1(S t) = e r(t k T 0 ) σk (S t + ν(t k t)) + µ k, Tk k=1 ξ 2 = T 0 t n k=1 e r(t k T 0 ) ξ 2 σ k Tk. ξ 2 28 / 53

29 Utility Maximization Assume exponential utility u(x) = e γx γ γ 0 represents managerial risk aversion Manager has two options: 1) invest in the market, or 2) invest in the real option Goal is to maximize the terminal utility in each of the two options and determine the indifference price 29 / 53

30 Optimal Investment in the Traded Index (Merton Model) Invest in market only, with π t invested in the risky asset dx t = (rx t + π t (µ r))dt + π t σdw t And maximize expected terminal utility V (t, x) = sup π t E [u(x T ) X t = x] Applying standard arguments leads to the PDE t V 1 (µ r) 2 ( x V ) 2 2 σ 2 xx V + rx xv = 0 with V (T, x) = u(x), and the solution is given by V (t, x) = 1 γ e 1 2( µ r σ ) 2 (T t) γe r(t t)x. 30 / 53

31 Optimal Investment in the Real Option Project Wealth dynamics are given as dx t = (rx t + π t (µ r)) dt + π t σdw t, t / [T 0, T 1,..., T n ] X T0 = X T K1 A 0 X Tj = X T + ϕ(s j )1 A, j [1, 2,..., n] j where 1 A represents the indicator function equal to 1 if the real option is exercised The manager seeks to maximize his expected terminal utility as U(t, x, s) = sup E [u(x T ) X t = x, S t = s] π t Applying standard arguments, it can be shown that the solution to U(t, x, s) can be achieved by solving the following PDE t U+rx x U+νs s U+ 1 2 ssuη 2 s 2 1 ((µ r) x U + ρσηs sx U) 2 2 σ 2 = 0 xx U 31 / 53

32 Optimal Investment in the Real Option Project (con t) Boundary conditions U(T j, x, s) = U(T + j, x, s)e γϕ(s), for j = 1,..., n 1 U(T n, x, s) = u(x + ϕ n (s)) Using the substitution U(t, x, s) = V (t, x)(h(t, s)) 1 1 ρ 2 results in the simplified PDE t H + ˆνs s H η2 s 2 ss H = 0 with H(T n, s) = e γ(1 ρ2 )ϕ n(s Tn ), and t (T n 1, T n ] Apply dynamic programming, where at each t = T j, j = {1, 2,..., n 1}, set H(T j, s) = H(T + j, s)e γ(1 ρ2 )ϕ j (S Tj ) 32 / 53

33 The Indifference Price At t = T 0, we should invest in the real option if (H(T + 0, s)) 1 1 ρ 2 e γker(tn T + 0 ) 1 Defining f as the indifference price, i.e. the value of the real option, and setting U(t, x f, s) = V (t, x) leads to f (t, s) = 1 ln H(t, γ(1 ρ 2 s)e r(tn t) ) 33 / 53

34 The Indifference Price for Normally Distributed Estimates Theorem Real Option Value of Risky Cash-Flows Accounting for Risk Aversion. For a given set of cash-flow estimates, normally distributed with mean µ k and standard deviation σ k, given at times T k, where k = 1, 2,..., n, the value of the option at time t < T 0 to invest the amount K at time T 0 < T k to receive these cash flows accounting for risk aversion, where the utility of the investor is given by u(x) = eγx γ, is given by f (t, s) = 1 ln H(t, γ(1 ρ 2 s)e r(tn t) ) where H(t, s) = Φ( B(t, s)) + e ξ2 t 2 Ĉ(t, s)φ(ξt B(t, s)). 34 / 53

35 The Indifference Price for Normally Distributed Estimates ξ t = γ(1 ρ 2 )â 1 T0 t n σ k â j = e r(tn Tk), bj = Tk k=j n µ k e r(tn T k) k=j ( ) γâj A j = â j 2 (1 ρ2 ) ν, A 0 = B(t, s) = n A j (T j T j 1 ) j=1 A 0 b 1 +Ke r(tn T 0 ) â 1 s ν(t 0 t) T0 t Ĉ(t, s) = e γ(1 ρ2 )(A 0 â 1 (s+ ν(t 0 t)) b 1 +Ke r(tn T 0 ) ) 35 / 53

36 Real Option Value (MMM) (a) Market Stochastic Driver as GBM (b) Market Stochastic Driver as GMR Project value and real option value of the UAV project for varying correlation (note that they are independent of S 0, ν and η) Correlation (ρ) Project Value (V 0) Option Value (RO 0) / 53

37 Sensitivity to Risk - Standard Approach Real Option Value (MAD Method) K = 40 K = 45 K = 50 K = Risk ( V ) 37 / 53

38 Sensitivity to Risk - MMM Assumptions: Single cash-flow at T 1 = 3 Expected value of the cash-flow: µ 1 = 50 Correlation to traded index: ρ = 0.5 Investment time: T 0 = 2 38 / 53

39 Sensitivity to Risk - MMM (ρ = 0.5) Real Option Value (RO 0 ) Risk ( ) K = 30 K = 40 K = 50 K = / 53

40 Sensitivity to Risk - MMM For a single cash-flow, the real option value is given as RO 0 = e rt 0 E Q ( ) e r(t 1 T 0 ) µ 1 + νσ 1 T1 }{{} Distorted Mean Recall ν = ρ µ r σ + e r(t 1 T 0 ) T0 σ 1Z T 1 } {{ } Standard Deviation + K 40 / 53

41 Sensitivity to Risk - MMM = 5 Frequency Distorted CF 41 / 53

42 Sensitivity to Risk - MMM = 5 = 25 Frequency Distorted CF 42 / 53

43 Sensitivity to Risk - MMM = 5 = 25 = 50 Frequency Distorted CF 43 / 53

44 Real Option Value - Indifference Price Project Price at S=S MMM ( = 0) = = = = (c) Real option indifference price as a function of S t and t at γ = Time (t) (d) Real option indifference price at S t = S 0 for varying levels of risk aversion. 44 / 53

45 Sensitivity to Risk - Indifference Price 45 / 53

46 Sensitivity to Risk - Indifference Price RO Indiff. Price RO Indiff. Price Risk ( V), = Risk ( V), = RO Indiff. Price RO Indiff. Price Risk ( V), = Risk ( V), = RO Indiff. Price Risk ( V), = / 53

47 Practical Implementation of the Matching Method Assume managers supply revenue and GM% estimates Scenario End of Year Sales / Margin Optimistic (50%) (60%) (65%) (60%) (60%) (55%) (55%) Most Likely (30%) (40%) (40%) (40%) (35%) (35%) (35%) Pessimistic (20%) (20%) (20%) (20%) (15%) (10%) (10%) SG&A* 10% 5% 5% 5% 5% 5% 5% Fixed Costs * Sales / General and Administrative Costs 47 / 53

48 Sales and GM% Stochastic Drivers Traded index di t = µi t dt + σi t dw t Sales stochastic driver to drive revenues dx t = ρ SI dw t + 1 ρ 2 SI dw t S GM% stochastic driver to drive GM% dy t = ρ SM dx t + 1 ρ 2 SM dw t M Cash flow V k (t) = (1 κ k )ϕ S k (X t)ϕ M k (Y t) α k 48 / 53

49 Bivariate Density of Sales and GM% Theorem The Bivariate Density of Sales and GM%. The bivariate probability density function between sales and GM% is given by ( u(s, m) =φ ΩρSM Φ 1 (F (s)), Φ 1 (G (m)) ) f (s) g (m) φ (Φ 1 (F (s))) φ (Φ 1 (G (m))) where φ Ωρ represents the standard bivariate normal PDF with correlation ρ, and φ is the standard normal PDF. 49 / 53

50 Project and Real Option Value Project value V T0 (X T0, Y T0 ) = Real option value n e r(t k T 0 ) E Q [v k (X Tk, Y Tk ) X T0, Y T0 ] k=1 RO t (X t, Y t ) = e r(t 0 t) E Q [ (V T0 (X T0, Y T0 ) K) + X t, Y t ] Risk-neutral measure ( ν = ρ SI µ r σ di t I t = r dt + σ dŵt, dx t = ν dt + ρ SI dŵt + 1 ρ 2 SI dŵ S t, and γ = ρ µ r ) SI ρ SM σ dy t = γ dt + ρ SI ρ SM dŵt + ρ SM 1 ρ 2 SI dŵ S t + 1 ρ 2 SM dŵ M t 50 / 53

51 Computing the Real Option Resulting PDE rh = H t + ν H x H + γ y H 2 x H 2 y 2 + ρ SM 2 H x y 40 Real Option Value ($) SI SM Value of the real option for varying ρ SI and ρ SM 51 / 53

52 Matching Method Conclusions The approach utilizes managerial cash-flow estimates The approach is theoretically consistent Provides a mechanism to account for systematic versus idiosyncratic risk Provides a mechanism to properly correlate cash-flows from period to period The approach requires little subjectivity with respect to parameter estimation The approach provides a missing link between practical estimation and theoretical frame-works 52 / 53

53 References Berk, J., R. Green, and V. Naik (2004). Valuation and return dynamics of new ventures. The Review of Financial Studies 17(1), Block, S. (2007). Are real options actually used in the real world? Engineering Economist 52(3), Borison, A. (2005). Real options analysis: Where are the emperor s clothes? Journal of Applied Corporate Finance 17(2), Henderson, V. (2007). Valuing the option to invest in an incomplete market. Mathematics and Financial Economics 1, Miao, J. and N. Wang (2007, December). Investment, consumption, and hedging under incomplete markets. Journal of Financial Economics 86(3), / 53