This chapter sets forth the Discussion Paper's analytical framework and research approach. It defines revenue risk, discusses various perspectives on such risk, and reviews the impact of revenue risk sharing on financeability. Furthermore, it discusses the trade-off between direct liabilities and contingent liabilities and the analytical framework criteria - value for money, financeability, fiscal impacts, and ease of implementation - that are used to evaluate the revenue risk sharing mechanisms. It also lays out the research approach, including literature review, case studies, interviews, and financial analysis. Finally, it provides context on how recent P3 transactions have been funded and financed.
The key drivers of revenue risk include: 1) traffic volume, 2) tolling regime (i.e., toll schedule) and 3) toll collection. Traffic and revenue (T&R) forecasts for many types of toll facilities - including greenfield toll roads, brownfield toll roads, and dynamically managed lanes - may have been too optimistic in early years or may have not considered the greater volatility. This "optimism bias" of Agencies, Developers, and all involved in both public and P3 toll road projects has been documented by several observers (Flyvbjerg et al, 2004 and 2005; Bain, 2009).
The uncertainty in T&R forecasts stems from a number of factors, including:
In addition, in some cases, Developers may have simply accepted optimistic forecasts to assist them in winning their bids. This may not be rational in the short-run but may yield long-run benefits from building up a portfolio of assets and experience in the U.S.
Since the beginning of the global financial crisis in 2007, vehicle miles travelled have declined and only in 2015 are they expected to surpass 2007 levels (USDOT FHWA OHPI, 2015). Researchers ascribe this unprecedented dip to a number of factors, especially the loss of jobs. However, some in the transportation community hold that the "millennial generation" is actively eschewing cars in much greater numbers than previous generations, seeking more urban housing and alternative commuting modes (American Public Transportation Association, 2013). Furthermore, with the rise of ride-sharing, shared mobility apps, automated vehicles, and telecommuting, some researchers such as David Levinson believe that there may be "structural disruptions" in the demand for road capacity and traditional car services in the next decades that make any type of T&R forecasting difficult (Levinson 2015). Credit rating Respondents said that, at the least, they severely discount growth rates in toll road forecasts beyond 25 years, as they understand that current known alternatives, such as public transportation, vehicle technology, and telecommuting could materially impact traffic growth rates.
Regardless of how these trends are impacting toll revenues in the short- and long-run it appears that Agencies, Developers, Equity Investors and Lenders will continue to view T&R forecasts as containing significant risk.
The Discussion Paper represents two major revenue risk viewpoints: 1) the Developer perspective, which includes Lenders and Equity Investors and 2) the public perspective of contracting Agencies and their constituencies, both which are discussed below.
The Developer perspectives on revenue risk can be seen from two separate viewpoints: 1) Developer/Equity Investors and 2) Lenders (including banks and bondholders). Developers/Equity Investors provide equity and are generally expecting to receive dividends in return. In a revenue risk toll concession, dividends can be highly volatile and may only be paid out after many years (if ever). Developers/Equity Investors bear the full upside/downside revenue risk and therefore expect a return that is commensurate with this risk.
Lenders on the other hand have no upside as they receive only interest on their loans. As a result, Lenders tend to be more conservative than Developers and impose stringent requirements to ensure they will receive interest and principal payments on time even in downside cases. Although Developers ultimately care most about equity returns, they know that without Lenders, the project cannot be cost-effectively financed. So their interests can be summarized as follows:
Since most P3s are financed primarily with debt, obtaining Lenders' approval is critical to successful P3s. A project's ability to attract non-recourse debt and equity financing is a key consideration when evaluating revenue risk sharing mechanisms.
While the Discussion Paper focuses on revenue risk sharing for P3s, the issues are similar in the transfer of revenue risk in tax-exempt toll road revenue bonds issued by Agencies, which can only pledge the toll revenues and reserve funds collateral. Bondholders of these non-recourse bonds, which are primarily U.S. individuals and some financial institutions, take on the risk that the Agency is able to repay on a timely basis or at all. While public toll road agency bankruptcies have been rare, changes in bonds' credit ratings have been more common, with such ratings affecting the price of those bonds on the secondary bond market (Tollroads, 2014).
Public toll road agencies may take similar measures as Developers to reduce the risk of revenue bonds in order to make them attractive to bondholders. These include increasing debt service coverage ratios by including more grants in the financing structure, increasing reserves, and/or adding subordinate debt (such as from the Transportation Infrastructure Finance and Innovation Act (TIFIA) program) to the financial structure. In particular, adding grants or increasing public funding (but not TIFIA loans) can be viewed as a form of public agency equity.
P3s are principally about efficiently managing risks, drawing on private sector expertise, and attracting private capital. From a VfM perspective, one of the key value drivers in P3s is optimal risk allocation between Agency and Developer. Optimal risk allocation means that a risk should be transferred to the party that is best able to manage that risk. However, the dilemma with revenue risk is that it is difficult to manage for both Agencies and Developers (Quiggin, 2005).
Although the Developer may be better positioned to provide customer service, manage certain revenue risks (such as toll collection technology and accident removal risks), and maintain the facility, other risks (such as population growth or traffic pattern changes) are largely beyond its control. Therefore, transferring all revenue risk to the Developer may be inefficient as Developers may struggle to price the risk efficiently (see Section 2.1.4).
Similarly, Agencies struggle to effectively manage revenue risk as they have limited influence over traffic. Most of the financial difficulties of recent U.S. and international P3 toll roads were due to ambitious projections of jobs and housing growth and the inability to weather economic cycles, including the global financial crisis, all of which are out of most Agencies' control. Managing national economic risks are primarily the responsibility of national governments', and is an enormous challenge.
However, Agencies may have control over a project's transport links, such as connecting interchanges and land use. They also have the ability to delay or accelerate the development of competing transportation, such as transit and parallel roads, and pursue economic development policies impacting toll revenues. These powers are rarely vested in one Agency, however, and differences among Agencies at various levels of government make unified policy decisions difficult. Nevertheless, it could be argued that Agencies are in a better, albeit far from perfect, position to manage revenue risk.
Based on the above, the VfM perspective is that it is more efficient for Agencies to retain most revenue risk. If an Agency is of the opinion that the Developer will perform better if exposed to revenue risk, Agencies could decide to transfer a small portion of the revenue risk to ensure that the Developer is sufficiently incentivized while still maintaining a high VfM.
However, VfM is only one perspective and ignores other considerations such as the fiscal impacts. From that perspective, revenue risk P3s are essential to finance infrastructure projects that would otherwise not be realized in the same time frame through public financing. This is because: 1) equity in a transaction increases the amount of financing capacity, as it is "patient" in early years, compared to traditional tax-exempt financings, 2) it may be easier for Developers to increase toll rates - subject to the concession agreement - than Agencies can. Furthermore, in the case of an AP concession, regardless of whether future toll revenues would cover AP payments and maintain an Agency's high credit rating, Agencies can reach a politically-determined debt limit. Given strong anti-debt movements throughout the U.S. and transportation's competition with other public priorities such as education and health care, it may be unrealistic for Agencies to develop all projects as AP concessions.
If the Agency transfers revenue risk to the Developer, the Developer faces the challenge of pricing this risk. From a VfM perspective, if the Developer cannot effectively manage this risk, it may either 1) conservatively price the revenue risk, potentially leading to inefficient pricing, or 2) decide not to bid, reducing the number of bidders, and possibly resulting in market failure.
To reduce the chance of market inefficiencies, revenue risk sharing mechanisms could be considered. In this case, the Agency still achieves some off-balance sheet financing while avoiding excessive pricing of risk or increasing the number of credible Developers willing to bid.
Based on the above concepts and a review of revenue risk sharing experience, the Discussion Paper suggests how Agencies may implement mechanisms on the continuum between AP at one extreme and a full revenue risk transfer to the Developer at the other.
When evaluating revenue risk sharing mechanisms, it is important to distinguish revenue risk from financial viability. Revenue risk refers to the uncertainty in the revenue stream. Typically, this uncertainty is reflected in the private cost of capital (debt and equity combined) associated with the project. If the net present value (NPV) of expected future cash flows discounted at the weighted average cost of capital (WACC) of the project is negative, a project is not financially viable.
One way to make such a project financially viable is to provide an upfront Developer subsidy, as has been common in many U.S. P3s. Providing such a subsidy to the Developer does not change the risk profile of future cash flows. However, with the upfront subsidy, the Developer is more willing to accept the revenue risks as the expected return on the Developer's capital is now equal to or exceeds the Developer's cost of capital. In other words, with an upfront subsidy, the Developer does not need to seek as much senior debt, can obtain cheaper senior debt, and/or can reduce the amount of the equity investment. Many of these benefits result in the Developer facing smaller potential losses from volatile cash flows. When these lower potential losses are combined with potential profits from other aspects of the project, such as the construction contract, the overall attractiveness of the project to the Developer increases.
Revenue risk sharing mechanisms do not necessarily replace or eliminate subsidies to financially unviable projects. However, they can reduce the required subsidy by changing the risk profile of future revenues. If the future revenues to the Developer become less uncertain - for example, due to a minimum revenue guarantee (MRG) - the required return on capital can be lower. As a result, a smaller upfront subsidy will be required to ensure that the expected future cash flows discounted at the lower project WACC is equal to or exceeds zero.
Another important distinction for Agencies when evaluating revenue risk sharing mechanisms is the difference between direct and contingent liabilities. An upfront subsidy or an AP is a direct liability; the Agency knows that it will incur this cost. A revenue guarantee, however, is a contingent liability. Depending on traffic and revenues, the Agency may or may not have to make a payment.
This uncertainty regarding the payment by the Agency to the Developer is an important feature of any guarantee. Unfortunately, this can also cause big surprises, as was the case in Portugal. With one of the world's largest P3 programs, it assumed significant contingent liabilities, resulting in a considerable fiscal burden in the aftermath of the financial crisis (Diu, 2014).
Furthermore, the uncertainty of contingent liabilities makes a fair comparison with direct liabilities difficult. For example, how does one compare a $100M upfront subsidy to a 20-year minimum revenue guarantee of $15M per year? If traffic is lower than expected, the guarantee could cost the Agency up to $300M in nominal value ($186M in NPV terms at a discount rate of 5%). However, if traffic were in line with expectations, it would mean that the Agency would not pay anything while still lowering the cost of capital of the Developer, due to reduced uncertainty of the revenue stream, and resulting in a lower bid. However, if the economy were to slow down, as in Portugal, the Agency's costs could be considerable.
As described below, contingent liabilities such as MRGs can help share revenue risk between Developers and Agencies. A number of academics have explored approaches to value contingent liabilities (Cheah and Liu, 2006; Chiara and Garvin, 2007; Irwin, 2007; Shan, Garvin, and Kumar, 2010). However, there is no single simple approach to undertake these valuations, making the exact cost to the Agency is difficult to determine.
Using the different perspectives, this Discussion Paper assesses the various revenue risk sharing mechanisms used around the world and in the U.S. More specifically, each revenue risk sharing mechanism will be evaluated using the following criteria:
Revenue risk sharing mechanisms are primarily about the allocation of a defined set of risks for a defined project. The mechanisms do not fundamentally change the project or its risks. This is why these mechanisms are not expected to have significant societal or welfare impacts, beyond the impacts already covered under the other considerations (including VfM and financeability). The mechanisms can, however, have an impact on the incentives of the Developer to maximize traffic or revenues, which could have welfare impacts. Even though the Discussion Paper does not primarily examine this broader economic perspective, it will point out the impacts of incentive mechanisms on welfare where appropriate.
Besides revenue risks, U.S. P3s are subject to numerous other risks, including:
The Discussion Paper does not address these issues. They are extensively reviewed in the project development literature. Nevertheless, these other risks can materially influence a project's risk profile and may affect their evaluation of overall risks. For example, a Developer may not bid on a toll road P3 with low revenue volatility if environmental issues and possible litigation have not been adequately addressed.
This section briefly describes the methodological approach used to review and evaluate different revenue risk sharing mechanisms.
The review of the literature on revenue risk sharing included:
Appendix V provides a listing of this literature.
The Discussion Paper's Appendix I describes several revenue risk sharing mechanism case studies that help to illustrate the mechanism and describe the context and evolution of these mechanisms. Based on these cases, mechanisms that are most appropriate for the U.S. market were selected for further analysis in Section 4.
More than 25 specialists involved in P3s were interviewed, including those employed at universities, research institutes, Agencies, Developers, Lenders, Equity Investors, law firms, and international finance institutions (i.e., the Respondents). They are listed in Appendix IV. To ensure that Respondents felt comfortable freely sharing their views, the Discussion Paper does not quote the Respondents directly, nor does it attribute any comments to any one entity. The exception is reference to the authors of published materials that are publicly available or Respondents who helped provide information on specific case studies.
For those mechanisms meriting U.S. consideration, a simplified financial model was developed to illustrate how each mechanism might apply and is used in Section 3. Some of the model parameters are derived from recent U.S. P3s, but have been simplified in order to convey how the mechanism would work. P3 transactions have many elements in common but differ enough that making generalized conclusions on the ideal mechanism can be difficult.
The Discussion Paper draws upon the U.S. P3 experience in the last two decades. Table 3 shows the financing structure typical of these transactions based on nine representative financings in which TIFIA has been involved (USDOT TIFIA, 2015).
| Financing Source | Revenue Risk P3 | Availability Payment P3 |
|---|---|---|
| Senior Debt | 20%-49% | 4%-40% |
| TIFIA Loan* | Up to 33% | Up to 33% |
| Equity/Deeply Subordinated Debt | 18%-47% | 6%-12% |
| Upfront Public Capital Contribution | 0%-33% | 20%-57% |
*A TIFIA Loan can cover up to 49% of eligible costs as defined by the TIFIA program, although no more than 33% has generally been provided; the actual TIFIA percentage may not necessarily reflect a percentage of the total financing structure, due to TIFIA's definitions of "eligible" costs.
Financing structures of many of these transactions are summarized on the USDOT TIFIA website (https://www.transportation.gov/tifia/projects-financed). These transactions frequently have four primary financing sources: senior debt, TIFIA loans, equity/deeply subordinated debt, and upfront public capital contributions, with the following characteristics:
Typically, transactions are structured to manage financing risk with mechanisms that address downside revenue scenarios. Structuring includes ensuring adequate debt service coverage ratios (i.e., a sufficient buffer to ensure debt service can be paid even in severe downside cases), debt service reserve funds, and/or ramp up reserves and other short-term liquidity facilities. Revenue risk P3s tend to have more equity than AP P3s, reflecting higher revenue volatility.
The "flow of funds" establishes the priority of payments as shown in Figure 1. Typically, toll revenues are used to first pay O&M costs, then senior debt, followed by the TIFIA loan, and finally equity. Actual flow of funds provisions are more complicated, including reserving monies to pay for hedging costs, reserve funds, and major maintenance. Financing sources higher in the waterfall are subject to less revenue and cash flow risk.
Figure 1: Example of Typical Flow of Funds or Cash Waterfall
Revenue risk is the risk that income at the "top of the cash waterfall" is not adequate to fund all "buckets" in the waterfall. The risk associated with financing is that the revenue available to pay senior and TIFIA debt is not adequate to meet those specific requirements. Financing risk incorporates both revenue risk and the risk that higher priority costs are greater than expected, e.g., higher O&M or higher major maintenance costs may absorb more cash than expected, leaving less for lower priority buckets. For most U.S. toll road P3s, revenue volatility is usually the most significant project risk.
Given the performance of past toll road transactions, Lenders have required new approaches to improve T&R forecasting. In particular, probabilistic analysis techniques that have been used in other infrastructure markets such as wind and hydropower are now being applied (see Figure 2). This method may better take into account the volatility of demand drivers, such as regional population and employment. Under a probabilistic approach, a Developer derives a 50% probability forecast (P50) as a base case. Under P50, actual revenues are expected to exceed the forecast 50% of the time. Based on Lenders requirements, the Developer and/or Lender typically also develop a downside forecast (P80 or P90), which has an 80% or 90% probability of being achieved, or a 10% or 20% probability of not being met.
Figure 2: Example of Probabilistic Approach to T&R Forecasts
View larger version of Figure 2
1 In the context of this White Paper, VfM assesses the extent to which a revenue risk sharing mechanism allows the Agency to receive (or pay) a fair price for the P3 contract. Ensuring a fair price implies avoiding any risk allocation that requires the Developer to inefficiently (and excessively) price the risk.
