[mathjax]
In this note, we provide some details and clarifications about our 2019Q4 index release with a focus on the performance of UK water utilities. Our 16th January release mentions that upward movements in gilt rates in the last months of 2019 led to negative returns for numerous infrastructure companies in that quarter, especially firms with very long-term cash flow forecasts, including utilities.
In our press release, we took the examples of Affinity Water and South East Water amongst the lower performers in our results for Q4. In effect, the whole UK water sector was systematically impacted negatively by rising rates. However, this effect varied between firms because of idiosyncratic (firm-specific) risks, in particular their exposure to interest rate risk as a result of
- firm-specific risk premia, and
- varying dividend growth forecasts.
Interestingly, those companies that were most impacted by the rate increase in Q4 also benefitted from the decrease of interest rates during 2019. So despite the challenges of the final quarter, they still turned out to be the best performers of the year.
In what follows, we return to our approach to compute the quarterly returns of unlisted infrastructure companies, especially large corporates that have very long-term investment horizons like UK water utilities.
Approach
We closely follow IFRS 13 guidance for the reporting of fair value:
- The unit of account is the whole firm and includes equity and quasi-equity (shareholder loans) i.e. any payouts that will accrue to the firm’s owners.
- We focus on ‘principal markets’ where trading is relatively common. For infrastructure firms, we consider countries in which the secondary market is at least 20% the size or volume of the primary market over the past two decades. The UK clearly qualifies, as does the utilities sector.
- We use the discounted cash flow (DCF) or ‘income’ method because direct and relevant comparisons (equivalent secondary market transactions) are too rare in private infrastructure markets.
- We use the latest available market inputs: for financials and cash flows, we rely on audited accounts and investor data (level-3), and for rates and risk premia on the latest market data (level-2, more on that below).
- We use this data to calibrate each valuation to the latest market information available at the time.
The gist of the IFRS 13 guidance and of our approach is that valuations should be as contemporaneous as possible in order to reflect current market conditions at the time of valuation, in this case, at the end of each quarter.
Raw data collection and cash flow forecasting
We first collect all available historical information about each of the 600+ firm that we track in our broad market index universe (25 countries). That includes all financials, events (debt refinancing, impairments, defaults, accidents, etc.), reconstituting the detailed financial structure of the firm instrument by instrument, etc. back to the incorporation date of the company or the project vehicle (SPV).
Based on historical financials and our analysis of each market, regulatory framework, etc. each company is then the object of a revenue growth forecast and debt service forecast made by the EDHECinfra team of financial analysts (and validated by three different humans), sometimes with inputs from the firm’s owners. For example, amongst the 12 UK water utilities that we track, the average revenue growth forecast for the next 25 years currently ranges from 0.4-1.6% per annum.
Source: EDHECinfra
We then proceed to forecast two quantities: the free cash flow of the firm (FCF) and its free cash flow to equity (FCFE) retention rate (RR).
Armed with 20 years of historical data for hundreds of companies and company-specific revenue and debt service forecasts, we run statistical models (using machine learning) of the systematic drivers and idiosyncratic trend of the free cash flow of each firm.
As this video explains ([popuppress id=”3095″]), historical data gives use a good understanding of how the free cash flow of the average infrastructure firm behaves given its revenues, debt service and various characteristics. We also forecast the idiosyncratic part of each firm’s free cash flow (the difference between the predicted average and actual cash flows). The combination of the systematic and firm-specific cash flow forecasts gives us a projection of the free cash flow process until the relevant horizon for each firm.
Likewise, we model each firm’s future free-cash flow to equity (free cash minus future debt service) retention rate, that is, its tendency to pay free cash flow to equity as dividends. The figure below shows the average 10-year payout ratio (defined as 1 minus the FCFE retention rate) for UK water companies.
Source: Companies House, EDHECinfra
The multiplication of forecast free cash to equity by the forecast payout ratio gives a long-range dividend forecast for each firm, that take into account the lifecycle of the firm, its specific characteristics (in particular how much debt it needs to pay) but also how infrastructure companies of the same TICCS® classes and subclasses tend to behave in terms of dividend payments.
Of course, such an exercise conducted over multiple decades is uncertain. But in combination with the free cash flow forecast, our retention rate modelling turns out to be a good predictor of future dividends. Utilities cash flows are also reasonable stable, and these models perform rather well and predict free cash several years ahead with very small out-of-sample errors.
These forecasts are repeated each quarter, using the latest available data (which is typically reported annually) and any known development in the country, sector or firm that may justify revisiting the revenue forecasts in particular in any given quarter.
For 2019, our medium-term dividend growth forecasts for UK water utilities ranged from close to zero to north of 6%, depending on our revenue growth assumptions, each firm’s current and future debt burden, and our free cash flow and retention rate projections.
Source: EDHECinfra
Estimating expected returns: mark-to-market discount rates
Once dividend forecasts are obtained, a discount rate is estimated for each company based on its individual risk factor loadings and the evolution of the price of each risk factor (or risk factor premia) in that quarter. Factor premia are derived from the secondary market for unlisted infrastructure equity. The discount rate (or expected return) $latex R_i$ of firm $latex i$ is computed according to the following formula (omitting the time period subscript for simplicity):
$$E(R_{i})=R_f+\lambda_{1} {\beta}_{i,1}+\dots+\lambda_{K}\beta_{i,K} = R_f+\sum_{k=1}^K \beta_{i,k}\times\lambda_k$$
We have established that for K=5 factors, a robust model of expected returns given the deal IRRs and risk premia that we have observed in secondary markets over the past 20 years can be calibrated. The five factors are the firm’s size (total assets), profits (return on assets before tax), leverage (senior liabilities to total assets), investment (capex to total assets) and the term spread for the relevant market. The model also includes control variables to capture effects specific to certain TICCS® segments like business model (here regulated firms) or certain sectors.
Each $latex \beta$ (beta) in the equation indicates a company-specific factor loading (a firm’s actual size, return on assets, etc. at one point in time) obtained from its latest accounts, and each $latex \lambda$ (lambda) represents a market risk premium re-estimated at the end of each quarter once new secondaries have been observed ([popuppress id=”3105″].)
The plots below describe some of the factor loadings of the firms we track in 2019Q4.
Source: Companies House, EDHECinfra
We estimate each risk factor premia by statistically decomposing the impact of factor loadings in actual secondary market transactions observed in each quarter over the past 20 years and updated (using more Machine Learning) each time a new transaction is observed. Hence, each quarter we compute the latest values of the size premium, the leverage premium, etc.
Combined with the latest term structure of “risk-free” rates (interpolated government bond yields at the relevant horizon), this firm-specific risk premia gives us a mark-to-market discount rate for each firm.
In the case of UK Water utilities in 2019, average estimated discount rates range from 5.92-6.71%.
2019 in the UK Water Sector
The dividend forecasts of all tracked firms are thus discounted using a company-specific, market-calibrated term structure of discount factors (one for each year/dividend until the investment end date), and the prices thus obtained can now be used to compute returns, cash yields, etc.
In the UK water sector, risk premia had trended slightly up during last year, affecting each firm as per their factor loading, but interest rates had decreased a lot, especially in Q3 which led to substantial capital gains on top of a healthy cash yield ranging from 2-9%.
In Q4 however, long term rates increased by about 40bp in the UK which increased discount rates across the board, impacting all UK Water Utilities negatively, differently depending on their individual risk premia.
Source: Datastream, EDHECinfra
We note that while South East and Affinity Water had the lowest quarterly returns in the final three months of 2019, they also have the highest one-year returns over the full year. In effect, given our dividend growth forecast, these two companies have a higher duration than their peers, and are more exposed to interest rate movements.
Their higher duration is due to their higher expected dividend growth starting from a lower level of dividend/revenues ratio than the rest of the sector, which can be higher than 20% in some cases.
A simple way to describe this phenomenon is to picture an equity stake in a water utility as a perpetuity with a rising coupon. The value of the perpetuity is given by:
$$PV = \frac{D}{ r-g}$$
where $latex PV$ is the price or present value, $latex D$ is the dividend/coupon at the time of valuation, $latex r$ is the discount rate and $latex g$ is the growth rate of dividends. This is of course the basis for the standard Gordon growth model.
If dividend growth is high relative to the discount rate i.e. if the growth adjusted discount factor is small, a change in the interest rates that make up $latex r$ more than proportionally impacts the discount factor. Say the discount rate is 6% (comprising a risk free rate of 1% and a risk premium of 5%) and dividend growth is 5%, giving a discount factor of 1%; any increase in the risk free rate of 50bp, which amounts to an increase of 8.4% of the discount rate $latex r$, results in an increase of the growth adjusted discount factor $latex r-g$ of 50%.. (assuming higher base rates do not impact dividend growth forecasts materially).
In the case of UK water utilities, durations range from 16-25 years, with the best performers for 2019 also exhibiting the highest duration in part due to their higher dividend growth forecasts. The first plot below shows individual quarterly returns for each company, highlighting the common impact of systematic risk factors and rate movements, as well as the differences between each firm.
The second plot shows the one-year total return (compounding quarterly total returns) for the same companies as well as the 1-year return of the EDHECinfra UK, Corporate Regulated Network Utilities Index for 2019 in red (Index Code: 99e45b90-c80e0b57). This index includes 28 firms in 2019, with weights of 48% for water companies, 15% for gas distribution companies, 5% for power transmission firms and 32% for power distribution companies. The index shows a 12.42% total return in 2019.
Water companies can thus be said to have outperformed in the UK regulated utility sector in 2019, primarily because of their greater exposure to certain systematic (and rewarded) risk factors including (but not only) interest rate risk.
Source: indices.edhecinfra.com
