From asset allocation to infrastructure investment

Transcription

1 From asset allocation to infrastructure investment 1/40 From asset allocation to infrastructure investment A roadmap for the development of institutional investment in infrastructure Frédéric Blanc-Brude, PhD Research Director EDHEC-Risk A presentation prepared for the NATIXIS European Infrastructure Day 17 October 2013, Paris

2 From asset allocation to infrastructure investment 2/40 Agenda 1 Is infrastructure an asset allocation question? 2 De ning relevant infrastructure instruments 3 Construction risk and debt portfolio construction 4 A roadmap to develop infrastructure investment solutions 5 First step: measuring the credit risk of unlisted infrastructure debt 6 Investment solutions: active creditor vs passive investor

3 From asset allocation to infrastructure investment 3/40 The world needs new infrastructure, but is it investors problem? Huge infrastructure investment needs have been identi ed Concerns about the sources of long-term nance $50Tr The OECD reports that The question of the role of institutional investors as providers of long-term nance to the real econommy is now on the high policy agenda Should investors care? Is there something special about investing in infrastructure? are needed by 2030

4 From asset allocation to infrastructure investment 4/40 Is infrastructure investment an asset-allocation decision? Asset allocation is institutional investors rst-order problem If infrastructure investment is relevant at the asset allocation level, it implies potentially large allocations: this is the link between the policy debate and investment decisions To answer this question we need an infrastructure investment benchmark ie robust expectations of what passive investment in infrastructure implies from a risk management point of view Is infrastructure investment new and unique compared to the existing investment opportunities?

5 From asset allocation to infrastructure investment 5/40 Is infrastructure investment an asset-allocation decision? The infrastructure investment narrative : Low price-elasticity of demand = low correlation with the business cycle Monopoly power = pricing power = in ation hedge Predictable and substantial free cash ow Attractive risk-adjusted cash yield, available over long periods Opportunity to invest in unlisted assets In other words, this ideal-type implies Improved diversi cation Better liability-hedging Less volatility than capital market valuations

6 From asset allocation to infrastructure investment 6/40 The wrong approach: infrastructure assets are not real Focusing on tangible infrastructure is the wrong approach: we are interested in nancial instruments and the risk factor exposures they create for investors Our proposed solution to address the public policy agenda (channelling institutional money towards infrastructure) within the prudential regulatory framework is to ignore the notion of infrastructure and instead to focus on project nance This approach greatly clari es the debate: 1 Project nance is well-de ned as a set of nancial instruments (cf Basel-2) 2 Project nance is a unique form of corporate governance and can be expected to create new and unique risk factor exposures compared to other types of underlyings 3 Project nance encompasses the immense majority of investable and stand-alone infrastructure projects in the world today and in all likelihood will be used to deliver most future projects (eg the various public-private partnership programs in Europe mostly imply project nancing according to the Basel-2 de nition)

7 From asset allocation to infrastructure investment 7/40 Relevant infrastructure investments Project nance: the art and science of manufacturing commitment to solve the perennial long-term investment problem 1995 & 2013 A double ltering mechanism: separate incorporation + leverage A unique type of corporate structure: a single-project rm with no other asset than receivables and a nite life between $3Tr worth of project nancing A structured approach: splitting the free cash ow of the rm by level of predictability: high leverage signals low asset risk A new perspective: project nance is not the median infrastructure project (that is a good thing!) Project nance is a unique form of corporate governance If contracts, not tangible assets, de ne the characteristics of infrastructure investments, project nance makes such contractual characteristics explicit

8 From asset allocation to infrastructure investment 8/40 Risk transfer and management in infrastructure project nance Contract with a public or private party creating a binding commitment to pay contracts with other firms to commission a number of tasks at a fixed price { #facility manager #operator #builder [rights to a pre-agreed revenue stream*] *subject to performance AND/OR [rights to a merchant revenue stream] License granting the right to operate a regulated monopoly on a commercial basis established by a consortium of firms to enter into a longterm contract {ProjectCo} raises long-term finance for the construction, operation and maintenance of the relevant infrastructure [rights to the firm s free cash flow] $senior debt$ $junior debt$ $equity$

9 From asset allocation to infrastructure investment 9/40 Project nance: Investing in base case cash ows Base case equity and subordinated debt service Base case senior debt service (2 facilities) Senior debt service cover ratio (DSCR) Example base case cash ows for a social infrastructure project nancing in the UK with a 30-year maturity The cash ow pro le is dynamic (predictable changes in time)

10 From asset allocation to infrastructure investment 10/40 Systematic risk factors and base case volatility base case volatility year 20 merchant revenues partly merchant year 5 pre-agreed revenues revenue risk model year 15 project lifecycle In the cross-section: contractual features explain cash ow volatility, especially revenue risk models (from guaranteed to commercial revenues) (Blanc-Brude and Strange, 2007; Blanc-Brude and Ismail, 2013b) Across the lifecycle: the impact of de-leveraging (Merton, 1974) Other controls: host-country factors

11 From asset allocation to infrastructure investment 11/40 Infrastructure project nance: the value proposition A [senior or junior]* xed term investment in the risky base case of a single-project rm that has contractual rights to receive an income and contractually managed capital and operating costs *delete as appropriate

12 From asset allocation to infrastructure investment 12/40 To build or not to build? The policy debate focuses on channeling long-term investors funds towards new infrastructure projects However, investors declared themselves mostly interested in existing projects exhibiting stable cash ows Is construction risk (the risk of construction cost overruns) too high for investors?

13 From asset allocation to infrastructure investment 13/40 What construction risk? Difference between ex ante and ex post construction costs (public) Infrastructure projects (see Flyvbjerg et al, 2004) n Mean 267% 33% Median 200% 00% Min -800% -230% Max 2800% 364% Std dev 550% 96% Skewness Kurtosis Source: Blanc-Brude and Makovsek (2013) Infrastructure project nance NATIXIS dataset Public construction risk - decision to build (Flyvbjerg dataset, n=110, ) Project finance construction risk - financial close (NATIXIS dataset, n=75, )

14 From asset allocation to infrastructure investment 14/40 So why is credit risk higher in earlier years? 35% 30% 25% 20% 15% 10% 05% 00% years of maturity project finance defaults Moody's A Moody s Baa Moody s Ba Observed marginal probability of default, 4,000+ project loans (Moody s, 2013)

15 From asset allocation to infrastructure investment 15/40 The dynamic nature of credit risk in project nance The leverage decision as an optimisation problem What is the max rm leverage achievable for a given level of credit risk? In project nancing, credit risk is engineered: the more predictable cash ows are, the more the asset can be leveraged Subsequently the rm either de-leverages or keeps its debt-to-asset ratio constant The impact of long maturities on credit risk can be offset by that of the continuous de-leveraging of the single-project rm (see Merton, 1974) This dynamic is capture by the evolutions of the debt service cover ratio: DSCR t = Cash Flow Available for Debt Service (CFADS) t Debt Service (Principal+Interest) t in each period t=1,2,t for a project nancing of maturity T

16 From asset allocation to infrastructure investment 16/40 The endogenous nature of credit risk in project nance In project nance, lenders have extensive control rights In the event of SPE default, lenders can use the project s cash ows beyond debt maturity as a form of collateral This is known as the loan s tail Loan recovery is therefore endogenously driven by the continuous and active involvement of lenders in monitoring, restructuring or extending failing loans Existing empirical evidence suggests 100% recovery of defaulted loans in two thirds of cases (see Moody s, 2013) We need to understand what this means from the perceptive of a passive investor aiming to capture the average risk-adjusted performance of infrastructure debt

17 From asset allocation to infrastructure investment 17/40 Debt portfolio construction: intuition Individual infrastructure debt instruments are expected to have a dynamic risk and return pro le (ie it changes over time) Performance is also expected to be driven by systematic risk factors in cross-section (between projects) We can treat different loans at different points in their lives as different assets to be combined into an ef cient debt portfolio, as long as: Different levels of credit risk attract corresponding levels of credit spreads Default correlations are low between loans at different stage in their maturity In this setting, we expect the earlier part of project loan s life (the green eld debt of new projects) to be a signi cant diversi er of the portfolio s risk

18 From asset allocation to infrastructure investment 18/40 Portfolio construction: basic ingredients adapting from Altman (1996) 1 Return measure EAR it = YTM it EL it for each risk factor i at time t in the loan s life Yield-to-maturity: we estimate the systematic drivers of credit spreads empirically Expected loss is EL it = PD it LGD it The Portfolio return measure is R p = N i=1 w iear it 2 Risk measure Assuming the binomial distribution of PDs, unexpected loss is UEL it = PD it (1 PD it ) LGD it The portfolio risk measure is written UEL p = N N i=1 j=1 w iw j UEL it UEL jt ρ ijt with w it, the weight of asset i in the portfolio at time from origination t and ρ ijt, the default correlation between asset i and asset j at time t 3 We build two portfolios: one with debt instruments drawn from the entire lifecycle including green eld debt, and one using only brown eld or post-construction loans

19 From asset allocation to infrastructure investment 19/40 Empirical inputs 1 Credit spreads Project nance credit spreads can be explained econometrically by loan, project and country risk factors (Blanc-Brude and Strange, 2007; Blanc-Brude and Ismail, 2013b) We use the predicted average spread level and time structure for a range of generic projects with different degrees of revenue risk, controlling for geography (Europe) 2 Credit risk The marginal PDs reported by Moody s (2013) can be tted as a logarithmic function of time, starting around 25% and trending down to almost zero by year 10 LGD is not very well documented and is xed at 20% 3 Correlations Like for corporate debt, default correlations are low and in part driven by the business cycle But the project lifecycle is shown to be a much bigger driver of default correlations projects are more likely to default in earlier years irrespective of the business cycle and mature project debt almost never defaults in bad states of the world It follows that default correlations are especially low across the project lifecycle Correlations across year from origination are reverse-engineered from Moody s (2013)

20 From asset allocation to infrastructure investment 20/40 So who is afraid of construction risk? Green elddebt-onlyportfolio Wholelifecyledebtportfolio PortfolioReturn (percentage) Brown elddebt-onlyportfolio Ef cient frontier and randomly selected portfolios using green eld ( rst 5 years), brown eld (last 15 years) and both (20 years) types of infrastructure debt (see Blanc-Brude and Ismail, 2013b) PortfolioRisk(percentage)

21 From asset allocation to infrastructure investment 21/40 Conclusion A portfolio of infrastructure project loans can be diversi ed across a number of systematic risk factors, in particular, revenue risk models which determine the volatility of the SPE s income stream, and the project lifecycle, which embodies the continuous de-leveraging of the SPE The multi-dimensionality of the risk and return pro le of infrastructure project debt suggests that several building blocks can be designed to create more ef cient portfolios The riskier and better remunerated early life of new project loans can be a useful addition to a basket of loans made to existing or brown eld projects We have learned more about the determinants of infrastructure project nance debt risk and return pro les but our results are limited by the quality and quantity of data available In particular, we have little or no way to predict credit risk based on observed defaults let alone model loss given default We must do better than that to benchmark infrastructure debt investments

22 From asset allocation to infrastructure investment 22/40 Measuring credit risk in project nance: a structural approach The dif culty with reduced form credit risk models: they rely heavily on historical data, while structural models require precise knowledge of underlying quantities (volatility) and a clear de nition of the default point + the many assumptions of the Merton model In the case of project nance debt, the structural approach is highly relevant: we can measure the volatility of the underlying and de ne the default point relatively easily using the debt service cover ratio of the borrower In project nance, the DSCR at time t is de ned as: DSCR t = Cash Flow Available for Debt Service (CFADS) t Debt Service (Principal+Interest) t Understanding what drives the volatility of DSCR t is both necessary and suf cient to document the credit risk of infrastructure debt

23 From asset allocation to infrastructure investment 23/40 De ning and predicting default Default is de ned as: Default t CFADS t < Debt Service (Principal+Interest) t Default t DSCR t CFADS t Debt Service (Principal+Interest) t < 1 In other words, the default point at time t is unambiguously DSCR t = 1 and the the probability of default p t at time t is written: p t = Pr(DSCR t < 1 min j<t DSCR j 1)

24 From asset allocation to infrastructure investment 24/40 DSCR and distance to default Without loss of generality, we can write (Crosbie and Bohn, 2003): Distance to Default = [Market value of assets] [Default point] [Market value of assets][asset volatility] In Blanc-Brude and Ismail (2013a), we show that distance to default can be written: DD t = 1 σ DSCRt Debt Service t 1 Debt Service t (1 1 DSCR t ) where σ DSCRt is the standard deviation of the annual percentage change in the DSCR value Hence, the distribution of DSCR t together with the debt repayment pro le (growth rate of the debt service) ie the debt base case, are suf cient inputs to estimate the Distance to Default of project nance loans

25 From asset allocation to infrastructure investment 25/40 DSCR and emergence from default Project nance debt is expected to have high recovery rates because the credit risk of individual borrowers is actively managed during the life of each loan In effect, lenders can use their extensive control rights to restructure or extend individual facilities, effectively lending against the balance of the projects cash ows (either during or after the loan s life) Hence, loans that default are also likely to emerge from default in the next period The distribution of DSCR t also captures this phenomenon: q t = Pr(DSCR t 1 DSCR t 1 < 1)

26 From asset allocation to infrastructure investment 26/40 Loss given default In Blanc-Brude and Ismail (2013a), we show that the loss at time t is written as the difference between the discounted base case debt cash ows and a factor which is the a function of the distribution of DSCR t with: w t = L t = B t 1 1+p t 1 q t /(1 p t 1 ) T i=t+1 D i (1 + r t ) (i t) (1 p i w i (1 E(DSCR i ))) Discounting is done using the debt yield-to-maturity at time t

27 From asset allocation to infrastructure investment 27/40 A Bayesian approach to establish the credit risk pro le of infrastructure debt Credit risk measures can be written as functions of the base case and the distribution of DSCR t The prior probability distribution of cash ows embodies an investor s subjective valuation and required rate of return { Ex ante cash ows: notional values The prior view Expected cash ows: likelihood to meet the base case We follow the same path: 1 Prior formulation: we formulate generic base case cash ows for infrastructure projects and propose to make an assumption about the probability distribution of the rm s free cash ow (cash ow available for debt service) over time 2 Simulated risk measures are produced using Monte Carlo simulation 3 Posterior formulation requires the collection of standardised data, as de ned by the methodology used to formulate the prior

28 From asset allocation to infrastructure investment 28/40 Prior formulation A generic economic infrastructure project with commercial revenue risk ie a growing and increasingly volatile DSCR t Variable Generic project Assumption CAPEX 100 Financial structure single equity and debt tranches SPE t 0 leverage 75% Debt ammortisation pro le constant at 6% int Debt Maturity 20 years Project Life 22 years DSCR at time 0 and T 13 and 16 Equity lock-up threshold DSCR=11 Prior assumptions Project free cash ow distribution Lognormal σ 2 DSCR σ 2 DSCR t+1 +01%

29 From asset allocation to infrastructure investment 29/40 Expected credit risk pro le Debt service cover ratio Average DSCR Probability of default 600% 500% 400% 300% 200% 100% Distance to default Probability of emergence from default 100% 99% 98% 97% 96% 95% 94% 000% %

30 From asset allocation to infrastructure investment 30/40 Expected credit risk pro le Distance-to-default and default frequency mapping Expected loss or loss given default Probability of default 600% 500% 400% 300% 200% y = 32297e -6193x R² = % 000% Distance to default 995% one-year value at risk 900% 800% 700% 600% 500% 400% 300% 200% 100% 00% % 450% 400% 350% 300% 250% 200% 150% 100% 050% 000%

31 From asset allocation to infrastructure investment 31/40 3 Data collection implications density 2 1 From an investor s perspective buying a basket of project loans, full knowledge of the distribution of DSCR t is suf cient to characterise the credit risk of infrastructure debt In combination with the debt base case (principal + interest) DSCR t captures lines relevant dimensions of asset value, asset volatility and the probability of reaching availpay the default point and to emerge from default merchant However, there are in all likelihood, several distributions of DSCRpartcont t, determined by each systematic risk factor driving the CFADS, in particular, Revenue risk factors 3 Counter party risks Other factors? econometric testing will be required in due course The main benchmarking objective is to document the distribution of DSCR t and its statistical determinants 0 density 2 lines availpay merchant partcont

32 From asset allocation to infrastructure investment 32/40 Conclusion: the roadmap We have established a methodology to measure risk in both senior and junior tranches of project nance investments: the main vehicle used to deliver infrastructure nancing This methodology relies on the parsimonious collection of standardised data about individual projects base case cash ows and the debt service cover ratio A number of steps remain to make long-term infrastructure investment relevant to institutional investors and identify the opportunity to modify the Standard Formula We call these steps the roadmap to develop infrastructure investment for long-term investors 1 Observed data will allow the formulation of a posterior probability distribution of DSCR t 2 The documented risk pro le of individual instruments (debt & equity) will allow the design of speci c, well-diversi ed investment benchmarks according to clearly de ned strategies and horizons 3 These benchmarks will reveal : whether the risk factor exposures created by infrastructure project nance are different from existing risk modules in the standard formula the correlation of project nance high losses with other states of the world the accessibility of investment solutions offering exposure to the risk factors identi ed

33 From asset allocation to infrastructure investment 33/40 The road map: making infrastructure relevant for investors

34 From asset allocation to infrastructure investment 34/40 Contribute! Developing institutional investment in infrastructure will not happen without a signi cant involvement from investors, managers, regulators and academics: 1 We need to think clearly about the mechanisms and instruments 2 We need to agree on methods, data reporting and benchmarking strategy 3 We need a lot more data! (to build our posterior view of probabilities) You can help (create a 5-trillion dollar industry) Read our papers Join the debate and support the development of adequate methodologies and a cash ow reporting standard for project nance Contribute your historic and future data

35 From asset allocation to infrastructure investment 35/40 The road map: progress to date

36 From asset allocation to infrastructure investment 36/40 Discussion: active creditor vs passive investor The past performance of infrastructure project nance debt is a consequence of its active management by lenders: managed credit risk What role will project lenders play in future investment solutions involving passive investors? The probability of emergence from default (and the size of recovery) of project loans can be measured by observing the distribution of DSCR t but they are explained by the behaviour of the lenders and their decision to restructure and reschedule existing debt when necessary The development of investment solutions accessible to the average investor implies transforming project debt into a product with a well-de ned horizon What role can loan tails or restructurings play in this context? Should green eld and brown eld debt be aggregated in separate funds which investors can combine? or invested together and managed dynamically? The possibility to improve portfolio performance by investing at different stages of the project lifecycle (green eld vs brown eld) also requires the ability to source and structure new green eld debt on an ongoing basis

37 From asset allocation to infrastructure investment 37/40 Even if you buy the best design long-term performance = maintenance Who will actively manage the credit risk of the infrastructure debt designed for passive institutional investors?

38 From asset allocation to infrastructure investment 38/40 Relevant publications: NATIXIS & EDHEC-Risk Research Chair on Infrastructure Debt Investment Who is afraid of construction risk? Infrastructure debt portfolio construction EDHEC-Risk Publications Frédéric Blanc-Brude & Omenia RH Ismail July 2013 Available at wwwedhec-riskcom/multistyle_multiclass/natixis_research_chair and cibnatixiscom/infrastructure

39 From asset allocation to infrastructure investment 39/40 Relevant publications: working papers Construction risk in project nance EDHEC Business School Working Paper Frédéric Blanc-Brude & Dejan Makovsek January 2013 Measuring the credit risk of unlisted infrastructure debt EDHEC Business School Working Paper Frédéric Blanc-Brude & Omenia RH Ismail August 2013

40 From asset allocation to infrastructure investment 40/40 References Altman, E (1996, October) Corporate Bond and Commercial Loan Portfolio Analysis Centre for Financial Institutions Working Papers 96-41, Wharton School Centre for Financial Institutions, University of Pennsylvania Blanc-Brude, F and O R H Ismail (2013a) Measuring unlisted infrastructure debt credit risk EDHEC-Risk Blanc-Brude, F and O R H Ismail (2013b) Who is afraid of Construction Risk? Portfolio Construction with Infrastructure Debt NATIXIS Research Chair on infrastructure debt investment EDHEC-Risk Blanc-Brude, F and D Makovsek (2013, January) Construction risk in infrastructure project nance Blanc-Brude, F and R Strange (2007) How Banks Price Loans to Public-Private Partnerships: Evidence from the European Markets Journal of Applied Corporate Finance 19(4), Crosbie, P and J Bohn (2003) Modeling default risk Technical report Flyvbjerg, B, M K Skamris Holm, and S L Buhl (2004) What causes cost overrun in transport infrastructure projects? Transport reviews 24(1), 3 18 Merton, R (1974) On the Pricing of Corporate Debt: The Risk Structure of Interest Rates The Journal of Finance 29, Moody s (2013, February) Default and recovery rates for project nance bank loans Technical report, Moody s Investor Service, London, UK