When investors need to restate liabilities – EDF

In measuring its €40bn French nuclear decommissioning liability, EDF applies a 10-year historical ‘sliding average’ discount rate to a current estimate of cash flows. In our view, this leads to an out of date (and at present understated) liability that you should not use in your analysis, even though the approach is deemed to comply with IFRS.

Smoothing out the effects of discount rate changes may reduce apparent volatility, but it does not help investors. Balance sheets should include realistic and fully up to date estimates of the present value of decommissioning and other similar obligations.


Decommissioning and environmental liabilities are significant for many companies, particularly those in the utility, energy and mining sectors. How these are measured impacts both performance metrics and the liability amount included in enterprise value calculations and DCF analysis.

The measurement of these liabilities essentially reflects the present value of cash flows expected to be paid to settle the obligation. Estimating those cash flows requires judgement and it is important to understand the basis on which forecasts are made and the sensitivity and risk factors involved. Liability duration is often long, with some cash flows not being paid for decades or more. This means that the value today is highly sensitive to the discount rate applied.

EDF total decommissioning liability €51bn, market capitalisation €39bn

For EDF the figures are particularly large given its extensive nuclear energy programme. At the end of 2018 the total balance sheet liability is €51bn and the undiscounted forecast cash flows on which this is based total well over €100bn. Considering that the market capitalisation of EDF is currently about €37bn, the measurement of the decommissioning obligation is clearly a key element for the analysis and valuation of the company.

Most of the liability (€39.8bn) relates to its French operations. We only focus on the French liability in this article as the other obligations, mainly those of EDF Energy, its UK subsidiary, are somewhat different in nature because they are covered by a receivable from the UK government. This means that the balance sheet valuation of the UK related liability is less important for investment analysis.

The chart below shows the total undiscounted expected cash flow and balance sheet liability for EDF’s French decommissioning obligations, together with the average liability duration estimated by us, using the sensitivity disclosures provided by the company. It is good to see EDF disclosing the undiscounted amount as this is helpful and not a specific IFRS requirement.

Undiscounted and present value of EDF decommissioning obligations in France
Source: EDF financial statements and The Footnotes Analyst estimates

We note that over this period the amount of undiscounted cash flows has increased by about €13bn and the time to when these are estimated to be paid has lengthened. The analysis of cash flow forecasts and their timing is important and EDF does provide extensive explanations and disaggregation of these amounts.

In this article we concentrate on the derivation of the balance sheet liability based on those forecasts. This part of the calculation is all about the discount rate.

Discounting of environmental provisions by EDF

For 2018 EDF uses a discount rate of 3.9% for its French nuclear decommissioning liabilities. The rate is based on the yield on bonds, with a duration similar to that of the liability cash flows and with the benchmark rate being identified as the yield on the French government bond OAT 4% Treasury Bonds 2055.

A rate of 3.9% is significantly higher than the risk-free rate of a similar duration.  At the end of 2018 the ECB quotes the Euro AAA spot rate for 22 years duration as 0.85%. There are two factors that explain the difference:

  • Rate averaging: Rather than take the rate at the balance sheet date, EDF uses a 10-year smoothed historical average rate. This averaging means that the rate applied is, at present, significantly higher than would be the case if based purely on the current rate.
  • Credit spread: EDF adds a credit spread to the government bond rate consistent with an A to AA credit rating, the rating it says is applicable to EDF itself. Also, our ECB spot rate reflects AAA Euro credit rather than the rates for the specific French government bonds used by EDF.

The company does not disclose the contribution made by each adjustment.

There is an additional factor that possibly explains the EDF rate. The company may be using a yield curve to discount each individual cash flow and not a single rate based on average cash flow duration. This is despite the company summarising the discounting process as a single rate in its disclosures. The use of a yield curve means that for long-dated cash flows the market rate would be unobservable. Choosing a high ‘ultimate forward rate’ in constructing that curve could lead to a higher average rate being applied than would be expected based on the average liability duration. In our calculations we have assumed that a yield curve has not been used. If it were (and we don’t know because there is no disclosure) then our estimated alternative measures would be impacted.

EDF decommissioning discount rate disclosure
Source: EDF 2018 financial statements

So, is the rate applied by EDF and the resulting liability relevant for investors? If not, what liability measure should you be using in your analysis?  To answer these questions we consider whether the rate applied by EDF is consistent with IFRS and comparable with that used for other similar liabilities.

IFRS discount rate requirements

Unfortunately, the IFRS accounting standard that applies to non-financial liabilities, IAS 37, is itself in need of updating. The requirements regarding the discount rate are somewhat vague and can, in practice, result in different interpretations. It says that the estimated cash flows should be discounted at a rate that reflects the “current market assessment of the time value of money and risks specific to the liability” (IAS 37, para. 47), but there is little additional guidance about exactly what this means, which undoubtedly contributes to the current diversity in practice.

The IASB’s interpretation committee discussed the IAS 37 discount rate in 2011, specifically the question of whether it should include a credit spread. The committee observed that the predominant practice is not to include a credit spread on the basis that such an adjustment does not reflect “risks specific to the liability”.  However, no formal interpretation or amendment to the standard was issued to make this a clear requirement and, based on our observations of EDF and RWE, different approaches clearly still persist.

Opinions differ regarding whether IAS 37 permits ‘sliding averages’

Whether the rate could be a historical average rather than a current rate at the balance sheet date is not directly addressed by IAS 37. Nevertheless, the reference to “current market assessments of the time value of money” seems to us to mean that the rate at the balance sheet date should be used.

Both PWC and EY reach the same conclusion in their IFRS reference manuals. EY says “ … this [an averaged rate] is clearly inappropriate given the requirements in IAS 37 to determine the best estimate of an obligation by reference to the expenditure required to settle it ‘at the end of the reporting period’, [IAS 37.36], and to determine the discount rate on the basis of ‘current market assessments’ of the time value of money [IAS 37.47]”

One could debate at length what is and is not compliant with IAS 37, but the flexibility in the standard is sufficient for Deloitte and KPMG, the joint auditors of EDF, to concur with the approach used by the company. Indeed, they specifically refer to the discount rate in their audit opinion and confirm its compliance.

“Concerning the inflation and discount rates adopted by management, we have verified their compliance with applicable accounting standards and regulatory measures, notably the ministerial order of March 21, 2007, as amended. We have reconciled the data used for this purpose with market data and available historical information.” Source: EDF 2018 financial statements, audit opinion.

If auditors consider that using an average rate is consistent with IAS 37 then, in our view, the standard needs to be changed and clarified. We do not believe that this type of smoothing or averaging provides relevant information for investors.

Furthermore, in our view discount rate averaging introduces an unnecessary accounting mismatch. EDF has a portfolio of dedicated assets that fund the decommissioning obligation, with most of these measured at current fair value. The effect of discount rate changes therefore has an immediate impact on the portfolio of financial assets but it has a delayed impact on the related liability. This inconsistency produces confusing metrics for investors.

While we are clear in our view that using a sliding average discount rate is not appropriate, the question of whether credit risk should be reflected in liability measurement is more complex. We are not going to get into the detail of this subject now, but merely emphasise that investors need comparability. An ‘observation’ by standard setters that the predominant practice is not to include credit risk has obviously not produced that comparability.

Comparison with other liability measurements

We compare the decommissioning discount rate applied in 2018 by EDF to two others. Firstly, the rate EDF itself uses to measure its own pension obligations and, secondly, the rate applied by German utility company RWE to measure its nuclear decommissioning liabilities.

Comparison with EDF pension liability

EDF has a second highly material liability in respect of pensions that, like decommissioning, is characterised by uncertain cash flows and long duration. The gross liability of its French pension schemes in 2018 is €29.2bn (€18.0bn is in the balance sheet as it is reported net of the fund assets). At 31 December 2018 the pension discount rate was 2.3%, significantly lower than the 3.9% applied to the decommissioning liabilities.

Considering the similarity of pension and decommissioning liabilities, one would expect some consistency in how these are measured. Clearly the nature of these liabilities differ in terms of their cash flow risks, credit risk and duration. However, it is difficult to see how these justify a such very different measurement basis.

The main reason for the difference in rate is that EDF applies a  ‘sliding-average’ approach in calculating the discount rate for its decommissioning liabilities but not in respect of pensions. This is because IAS 19, the IFRS standard that applies to pensions, is more prescriptive than IAS 37 and explicitly rules out any such discount rate averaging.

Comparison with RWE decommissioning liability

German utility company RWE also operates nuclear power stations and has a decommissioning liability in its balance sheet. There is an important difference compared with EDF, which is that in Germany the German state is responsible for parts of the longer-term decommissioning costs. This results in a relatively smaller liability with a shorter duration (we estimate under 10 years).

The shorter duration partly explains the lower discount rate of 0.4%. However, the dominant reason for the low rate is that RWE does not appear to use any form of averaging or include a credit spread. Its discount rate is not much more than the 10-year AAA spot rate quoted by the ECB for December 2018 of 0.32%. The lack of comparability due to EDF and RWE using such very different approaches to discounting is not helpful for investors

The chart below summarises the various rates we highlight above.

Alternative discount rates
Source: EDF and RWE financial statements and ECB published yield curves. We have not shown RWE prior to 2016 as it used a very different approach to discounting. It applied a rate similar to that used by EDF, but with the effect mitigated by using a higher risk adjusted cash flow escalation rate.


Recalculating the EDF liability at different discount rates

As we said above, liability measurement matters in equity analysis and it is important that where such liabilities are material you question the methodology and assumptions. Where necessary you should recalculate the liability at rates you feel are more appropriate. 

Using sensitivities to estimate alternative liability measures may not be accurate

It is not possible for investors to accurately adjust balance sheet liability measures for alternative discount rates. To do so would require knowledge of the full schedule of forecast cash flows. But it is possible to make an approximation based on the sensitivity or duration disclosures. All companies are required to explain the sensitivity of liability measures to a given change in discount rate. EDF discloses the effect of a 0.2% reduction and 0.2% increase in rates; these produce an increase of €1.8bn and reduction of €1.6bn in the liability respectively. Sensitivity reflects the duration of the cash flows – longer duration results in higher sensitivity.

For small changes in rates, using these sensitivities is fine. But for larger changes extrapolating becomes increasingly inaccurate. The reason is that the change in present value is not linear. As can be seen from the data above, sensitivity to rate changes is higher at lower discount rates, an effect called convexity in financial mathematics. We have allowed for the estimated effect of this in our calculations below.

For the period 2014 to 2018 we recalculated an estimated liability using both the rate EDF uses for its own pension obligations and the ECB AAA rate for a duration equal to our estimate of the EDF decommissioning liability duration. The use of the AAA rate is consistent with the approach we believe is used by RWE, although the rate is higher owing to the longer duration of the EDF liability.

As one would expect, given the high sensitivity to different discount rates, the liabilities are very different. Notice that using current rather than 10-year averages also increases the volatility of the liability.

EDF French decommissioning liability estimated using different discount rates
Source: EDF financial statements and The Footnotes Analyst estimates

So what liability measure should you use?

Averaging discount rates and smoothing of value changes does not help investors. We do not agree with the reasoning of EDF that a sliding 10-year average rate in some way “prioritises long-term changes in rates” or that it is justified by the long-term nature of the cash flows. The fact that rates will change in future periods does not mean that current interest rates are somehow inferior to out of date averages.

We think that where companies have used an averaged rather than current discount rate in measuring liabilities, you should not use the balance sheet figure in your analysis and valuations. Instead, you should adjust to an estimated current value. Credit risk is more complicated – we are of the view (others will likely disagree) that an economic measure of any liability should include applicable credit risk, but this should take into account the characteristics of the liability, particularly the existence of a dedicated investment fund and the resulting credit risk mitigation.

In our view the EDF French decommissioning liability relevant to investors is higher than the amount of €39.8bn reported, under IAS 37, in its balance sheet. We would use something between the €53.5bn we estimated by applying the IAS 19 rate and the €69.8bn we estimated using the ECB 22-year AAA rate.  An amount closer to the lower figure is probably most appropriate considering our belief that the liability should incorporate the effect of credit risk.