EDHEC-Risk Days Europe 2015

Transcription

1 EDHEC-Risk Days Europe 2015 Bringing Research Insights to Institutional Investment Professionals Mars The Brewery - London The valuation of privately-held infrastructure equity investments: Theoretical framework and data collection requirements Frédéric Blanc-Brude, PhD Research Director EDHEC-Risk Institute

2 EDHEC-Risk Days Europe 2015 Agenda Bringing research insights to institutional investment professionals 1 Objectives 2 Approach 3 Expected dividend model 4 Discount rate term structure model 5 Example 6 Data collection 7 Next steps

3 Objectives Benchmarks of privately-held infrastructure equity investments aim to: 1 provide measures of realised returns and risk requires computing an end-of-period value 2 allow forward-looking measures of risk-adjusted performance, including: expected returns, risk and correlations (for asset allocation) extreme risk measures eg VaR (for prudential regulation) duration (Infrastructure and LDI) Current tools and approaches only allow addressing the realised performance measures but rather imperfectly and do not address any of the forward looking questions

4 Existing approaches 1 Listed proxies 2 Public-market equivalents 3 Repeat sales and other real estate methods 4 IRRs and appraisal-based NAVs

5 Existing approaches: Listed proxies In a forthcoming paper (Blanc-Brude et al, 2015), we show that adding listed infrastructure to the portfolio of a large, well-diversified investor (by asset class or by factor exposures) fails to improve the investable universe of efficient portfolios (the mean-variance efficient frontier) Efficient Frontier July 2000 to December Returns 005 Hedge Funds Fixed Interest FTSE Macquarie Infra Developed Equities Real Estate Emerging Equities Commodities Standard Deviation Reference Assets Reference and Infrastructure Assets

6 Existing approaches: PMEs 1 Index-based public market equivalent: investing the cash flows of private investment in and out of an public market index of choice (Ljungqvist and Richardson, 2003; Kaplan and Schoar, 2005) Assumes a beta of 1 Assuming betas is problematic and is likely to over/under-estimate true outperformance 2 Industry-beta PME: matching private investments with listed industry betas (Kaplan and Ruback, 1995; Phalippou and Zollo, 2005) Un-leverage the betas and re-leverage them with investment specific beta Requires knowing the beta of the debt of the private project We already know that matching infrastructure investments with industrial sectors is not straightforward This approach assumes a constant equity beta, which is counter-intuitive with greenfield infrastructure projects

7 Existing approaches: Repeat sales Developed in the real estate sector Has been applied to private equity investments (venture capital, buy-outs) and hedge funds (Woodward, 2004; Cochrane, 2005) financing rounds and IPOs are used as infrequent but multiple market valuations But the frequency of secondary transactions in infrastructure projects is unlikely to be high enough to allow this kind of approach Stale pricing issue (serial correlation of returns) Sample bias because only the better venture are financed multiple times No space for subjective investor valuations

8 Existing approaches: IRRs and appraisal-based NAVs 1 Issues with IRRs Single-rates are not adequate when investment projects has several phases (Brealey and Myers, 2014) - instead they imply that longer cash flows are riskier A single rate gives the accurate price only if the value is already known There are many other quibbles with IRRs (cash flows switching signs, beta identification, etc) Even if the correct value is known, a constant IRR creates other issues: duration cannot be calculated accurately, beyond one asset, it is not possible to average IRRs to get a portfolio measure, IRRs do not allow factor decomposition in each period (sources of return), etc 2 Issues with reported NAVs Different valuation methods make reported NAVs difficult to compare directly The literature documents systematically opportunistic NAVs reporting by managers (PE, hedge funds etc) (see for example Jenkinson et al, 2013, for a recent empirical study)

9 Proposed approach Current methods used by long-term investors (including direct investors) in infrastructure are not very satisfactory when it comes to realised performance and do not provide much in term of forward looking measures To compute a value for a privately-held infrastructure equity investment, we have to discount a stream of future cash flows (as for any other stock!) However, we have to take into account the characteristics of private assets, in particular, cash flow volatility and discount rate volatility have to be treated as separate (albeit related) phenomena: Private markets are incomplete There is no unique price for the same infrastructure investment What the market does implicitly for public stocks (aggregate all available information to discount a future stream of dividends) needs to be done explicitly for privately-held infrastructure equity stakes: 1 A model of expected dividends 2 A model of the term structure of discount rates (expected returns)

10 Proposed approach: Intuition If we can document the distribution of future dividends for a given risk profile of infrastructure investment and we can observe individual investors initial equity (price) in a given profile then as long as: 1 the terminal value is either zero (most cases) or known, and 2 the volatility of dividends at each point in the future can be assumed to justify the level of expected period returns of the average investor we can infer average expected returns over the life of the investment, as well as the range of expected returns implied by the initial range of prices

11 Expected dividend model Challenges: limited availability of cash flow data in the cross-section (between projects) and in time (most projects are too young) Bayesian response: 1 Group projects in a priori families of risk profiles based on contractual and financial characteristics (existing information) 2 Learn as much as possible about the true underlying cash flow distribution through observable data Keep the project categorisation (by family) simple eg 2 2 matrix: revenue risk [high vs low] and lifecycle [greenfield vs brownfield] Improve and develop the cash flow model as we learn and expend the database Use the dividend base case as a starting point to derive a dividend distribution for the entire lifecycle (requires estimating the parameters of the distribution of the ratio of realised dividends to base case dividends for each family of projects)

12 Expected dividend model The equity service cover ratio We use the investment base case as a point of reference to compute the Equity Service Cover Ratio For a stream of cash flow to equity C i t in each future state of the world i at time t: ESCR i t = Ci t C 0 t E t (ESCR t ) provides a direct measure of expected cash flows when multiplied by the base case dividends of a given project σ t ESCR t is a direct measure of cash flow volatility The output of our cash flow model is the conditional distribution of ESCR t

13 Expected dividend model Discovering families of ESCR distributions We model/calibrate two dimensions of the dividend stream process: 1 The likelihood of getting paid (any amount) at a given point in time given what occurred at the previous period (Markov Chain) 2 The distribution of dividends when/if investors get paid Using Bayesian inference allows updating/improving a prior view on these numbers as more data is observed In this paper, we show that even if the prior is very far from the true value, estimates converge quickly after only a dozen observations We calibrate several families of ESCR distributions 1 Merchant vs Contracted projects 2 Different financial structures 3 Different stages in the project lifecycle

14 Calibrating ESCR disitributions State transition probabilities Distribution of ESCR in the payment state round 1 true value round 1 round 2 prior estimate true density posterior estimate true value round 2 round 3 prior estimate true density posterior estimate true value round round 4 true value round 7 true value round 4 prior estimate true density round 5 posterior estimate true value round 7 prior estimate true density round 8 posterior estimate true value ESCR Density round 6 true value t= t= t= round 5 round 8 round 9 prior estimate true density posterior estimate Merchant Infrastructure prior estimate true density posterior estimate true value ESCR round 00 0 ESCR Density 2 1 round

15 Term structure model Challenges: the expected returns of individual investors are not revealed by building (a portfolio of) traded instruments expected returns (discount rates) are not unique and are unobservable But given an initial equity investment value (observable) and the statistical distribution of future dividends (also observable), we can extract the implied average term structure of expected returns and its upper and lower bounds Simple implementation: state-space model and Kalman filter The state-space model presented in the paper uses a dynamic Gordon model of discounted dividends (measurement equation) and a single-factor, auto-regressive term structure model of discount rates (state equation) Standard Kalman filtering implied gaussian error terms, but this can be relaxed It could be more complex is the data requires it (Monte Carlo Markov Chain) Derives the expected returns that best match the price paid by investors for a stream of future cash flows, assuming that they price the risk (conditional volatility ) of future dividends

16 Term structure model State-space model Observation equation: T P t0 = m f t o,t c C 0 t c+τ E t0 (m tc +τ ESCR tc +τ ) + ϵ 1,t τ=1 (unobservable) State equation: with m t+τ = e τµt+τ and µ t+τ = λ i t = ϕλ i t 1 + γ t σ i t + ϵ 2,t τ 1 i=0 (rfree t+i +λ t+i) τ

17 Term structure model Filtering implied discount rates ˆ + ˆ ˆ + ˆ ˆ ˆ

18 ESCR t = Dividend t Base Case Dividend t range of investments (prices) in the same cash flow process dividend distribution at time t: ESCR t and σescr t t 0 t 1 t 2 t 3 t 4 t 5 t 6 State-Space Model observation equation: P t0 = m f to,tc T τ=1 C0 tc+τ Et0 (m tc+τ ESCR tc+τ ) + ϵ 1,t state equation: λ i t = ϕλ i t 1 + γ t σ i t + ϵ 2,t Kalman Filter Range of implied term structures Term Structure (%) 15 Mean 95% ci bounds time Evolution of implied term structure

19 Example: Dividend distribution & expected returns Base Case & Expected Dividend Dividend Volatility 10 8 Base Case Expected Dividend (k$) 6 4 Volatility (%) Period Period Expected Single Period Returns Expected Multi Period Returns (Discount Rates) 25 Mean Return 20 BC IRR std Bounds 15 Mean 1 std Bounds rt+τ(%) 10 µt+τ(%) Period Period

20 Example: Valuation and implied market prices Stochastic Discount Factor Equity Price 10 Mean 15 Mean 08 1 std Bounds 1 std Bounds mt+τ Price (k$) Period Period Evolution of IRR Evolution of Prices 16 BC IRR 12 Exact Mean IRR (%) Filtered IRR 1 std bounds Price (k$) Observations Filtered Mean 1 std Bounds Deals Deals

21 Results Now we can compute Realised returns (conditional on end of period valuation) Current value (based on the latest expected cash flow stream and the implied term structure of discount rates) Expected returns (for asset allocation) Value-at-Risk (for prudential models) Duration (for ALM)

22 Data collection The paper aims to minimise the amount of data that needs to be collected from investors to be in a position to deliver our objective metrics The required data can be collected from investors and consists of 1 Data available at financial close (project dates, base case cash flows, characteristics) 2 Data collected on an ongoing basis by investors and their agents (realised dividends, state of the investment at each point in time) The data collection template is both parsimonious and realistic and sufficient to implement the approach proposed

23 Conclusions With this paper we have established: A technology for pricing privately held infrastructure assets, understanding private market dynamics and deriving forward-looking measures of performance and extreme risk A comprehensive (and efficient) data collection initiative A standardisation framework for data and performance reporting

24 Conclusions: the roadmap These results are part of a larger workplan initiated in 2012 with the Meridiam/Campbell Lutyens research chair at EDHEC 1 Identify relevant investable infrastructure assets 2 Engage the industry and regulators about the nature and drivers of performance of infrastructure investment 3 Design adequate valuation methodologies 4 Determine data collection requirements 5 Launch a global data collection initiative 6 Measure volatilities and correlations 7 Develop insights into portfolio construction with illiquid assets 8 Implement these findings in coordination with the industry to produce useful measures of performance for asset allocation and prudential regulation purposes

25 Next steps: 2015 The third paper of the Meridiam/Campbell Lutyens research chair aims to document the distributions (especially the volatilities) and the correlations of different groups of infrastructure dividend streams It involves collecting data spanning the past years in the OECD, combining merchant and contacted (PPP) projects It will be instrumental to answer EIOPA s current question: What is the likelihood of making large losses from infrastructure equity investments in bad states of the world?

26 MERIDIAM & Campbell Lutyens Chair The valuation of privately-held infrastructure equity investments Theoretical framework and data collection requirements, EDHEC-Risk Institute Publications Institute by Frédéric Blanc-Brude & Majid Hasan March 2015

27 References Blanc-Brude, F, S Wilde, and T Whittaker (2015) The performance of listed infrastructure equity: a mean-variance spanning approach Brealey, R A and S C Myers (2014) Principles of Corporate Finance (eleventh ed) McGraw Hill Cochrane, J H (2005, January) The risk and return of venture capital Journal of Financial Economics 75(1), 3 52 Jenkinson, T, M Sousa, and R Stucke (2013) How Fair are the Valuations of Private Equity Funds? SSRN Electronic Journal Kaplan, S N and R S Ruback (1995) The valuation of cash flow forecasts: An empirical analysis The Journal of Finance 50, Kaplan, S N and A Schoar (2005) Private Equity Performance: Returns, Persistence, and Capital Flows The Journal of Finance 60(4), Ljungqvist, A and M Richardson (2003) The cash flow, return and risk characteristics of private equity Working Paper, 1 43 Phalippou, L and M Zollo (2005) What drives private equity fund performance Unpublished working paper Woodward, S E (2004, August) Measuring Risk and Performance for Private Equity