Strengths, limitations and the M&A processes in which both valuation approaches are best combined.
The debate between discounted cash flow (DCF) and comparable multiples is framed the wrong way. A multiple is not an alternative to a DCF: it is a DCF with all its assumptions compressed into a single number and borrowed from other companies. This article proves this formally and draws the consequences. It explains when each method fails, what the empirical evidence says and how to combine them in an M&A process so that each one checks the other’s errors.
Every multiple is a compressed DCF
An EV/EBIT multiple depends on only three variables: the return on new investment, growth and the cost of capital. This can be shown in four steps.
1. Free cash flow and reinvestment. Free cash flow is net operating profit after tax (NOPAT) minus net investment I. To grow at a rate g, the company invests a fraction of its NOPAT that depends on the return on new invested capital (RONIC):
I_1 = \frac{g}{\mathrm{RONIC}}\,\mathrm{NOPAT}_1 \quad\Rightarrow\quad \mathrm{FCF}_1 = \mathrm{NOPAT}_1\left(1-\frac{g}{\mathrm{RONIC}}\right)
2. Value with constant growth. With a WACC greater than g, enterprise value is a growing perpetuity (Gordon):
\mathrm{EV}_0 = \frac{\mathrm{FCF}_1}{\mathrm{WACC}-g} = \frac{\mathrm{NOPAT}_1\left(1-\frac{g}{\mathrm{RONIC}}\right)}{\mathrm{WACC}-g}
3. The multiple. Since NOPAT = EBIT × (1 − t), dividing by next year’s EBIT:
\frac{\mathrm{EV}_0}{\mathrm{EBIT}_1} = \frac{(1-t)\left(1-\frac{g}{\mathrm{RONIC}}\right)}{\mathrm{WACC}-g}
This is the so-called value driver formula popularised by Koller, Goedhart and Wessels (McKinsey). Its consequences are less well known than the formula itself.
4. Corollary: growth is not always worth paying for. Differentiating the multiple with respect to g:
\frac{\partial}{\partial g}\left[\frac{1-g/\mathrm{RONIC}}{\mathrm{WACC}-g}\right] = \frac{1-\mathrm{WACC}/\mathrm{RONIC}}{(\mathrm{WACC}-g)^2}
The denominator is always positive, so the sign is determined by RONIC relative to WACC. If RONIC = WACC, the multiple equals (1 − t)/WACC regardless of growth: growing does not create value. If RONIC < WACC, growing destroys it.
With t = 25% (the standard corporate income tax rate in Spain) and WACC = 9%:
|
RONIC |
EV/EBIT with g = 1% |
g = 2% |
g = 3% |
|---|---|---|---|
|
9% |
8.3x |
8.3x |
8.3x |
|
15% |
8.8x |
9.3x |
10.0x |
|
30% |
9.1x |
10.0x |
11.3x |
The practical consequence is straightforward. Applying the median EV/EBIT of a peer group is equivalent to assuming that the company being valued has the same marginal return, the same long-term growth and the same cost of capital as that median. Anyone who uses multiples “to avoid making assumptions” is making the same three assumptions as a DCF, just without seeing them.
The fragility of the DCF: almost all the value lies at infinity
A DCF makes its assumptions explicit, and that is its great strength. Its weakness is that the result depends in a highly non-linear way on two of them.
The duration of value. In the Gordon formula, the semi-elasticity of value with respect to WACC is exactly:
\frac{\partial \ln \mathrm{EV}}{\partial\, \mathrm{WACC}} = -\frac{1}{\mathrm{WACC}-g}
With WACC = 9% and g = 2%, that “duration” is 14.3 years. One additional point of WACC reduces value by around 14%, to a first approximation. And because the denominator is a difference, the effect grows explosively as g approaches WACC.
An example. A company with free cash flow of 100 growing at 5% for five years and then at 2% in perpetuity, discounted at 9%, is worth 1,656. 73% of that value is terminal value: cash flows beyond year five. Change in value relative to that base case:
|
WACC |
g = 1.5% |
g = 2.0% |
g = 2.5% |
|---|---|---|---|
|
8% |
+9.7% |
+16.9% |
+25.5% |
|
9% |
−5.2% |
0 |
+6.0% |
|
10% |
−16.6% |
−12.7% |
−8.3% |
A range of ±1 point of WACC and ±0.5 points of growth, perfectly defensible before any expert, produces values between −17% and +26%. That is a spread of more than 40 points on reasonable assumptions.
The most common error is not the rate, but consistency. Many models calculate terminal value using the final-year cash flow growing at the chosen g, without adjusting for the reinvestment needed to fund that growth. According to the formula in the previous section, this is equivalent to assuming an infinite RONIC: growing for free, forever. It is an upward bias that cannot be seen in the spreadsheet and is only spotted by those who know the identity. Two further consistency checks: g must be nominal if WACC is nominal, and it cannot exceed the nominal growth of the economy on a sustained basis.
The fragility of multiples: errors you cannot see
Multiples have a virtue the DCF lacks: they capture what the market is paying today. That is precisely why their errors also matter. Four are especially common.
1. How they are averaged. The arithmetic mean of peers’ P/E ratios systematically overstates value. By the inequality of means (a consequence of Jensen’s inequality, since 1/x is convex):
\frac{1}{n}\sum_{i=1}^{n}\frac{P_i}{E_i} \;\geq\; \left(\prod_{i=1}^{n}\frac{P_i}{E_i}\right)^{1/n} \;\geq\; \frac{n}{\sum_{i=1}^{n}E_i/P_i}
Equality only holds when all the multiples are identical. With three peers at 8x, 12x and 30x, the mean is 16.7x, the median 12.0x and the harmonic mean 12.4x. The harmonic mean is equivalent to averaging yields (E/P), which is the economically meaningful quantity. A single peer with depressed earnings drives up the arithmetic mean and barely moves the harmonic mean. Liu, Nissim and Thomas (2002) confirmed empirically that the harmonic mean produces smaller valuation errors than the mean and the median.
2. Which multiple. In the same study, forecast earnings (forward P/E) were the best value driver in 66 of 81 industries. The interquartile range of errors was 0.32, compared with 0.74 for sales. Enterprise value to sales and to EBITDA multiples lagged behind. This is uncomfortable evidence for standard mid-market practice, which values almost everything on EV/EBITDA. With one important caveat: the sample excludes loss-making companies, precisely where earnings multiples are of no use.
3. IFRS 16 versus Spanish GAAP (PGC): the Spanish error. Since 2019, listed companies applying IFRS 16 recognise almost all leases as an asset and a liability. Their EBITDA no longer includes rental expense. A Spanish SME applying the PGC still records it as an operating expense. Applying a listed company’s EV/EBITDA to the EBITDA of an unlisted company mixes two different definitions. A Duff & Phelps study of the IBEX 35 found that the effect of IFRS 16 on the multiple exceeded 0.1x in 46% of cases, and in some exceeded 2.0x. The most affected sectors were retail, hotels, telecoms and airlines. Figures must be made consistent: either adjust the company’s EBITDA and add lease liabilities, or restate the peers on a pre-IFRS 16 basis. Those reporting under US GAAP (ASC 842) require a third treatment.
4. Which level of value is being imported. A listed company’s multiple measures the price of a liquid minority stake. A transaction multiple includes a control premium and the synergies paid for by a specific buyer. It also suffers from the winner’s curse: in an auction, the winner is, on average, the bidder who most overestimated the value. Neither directly measures an interest in an unlisted company, which requires adjustments for illiquidity, size and control.
What the evidence and the experts say
The empirical evidence does not produce a winner. It shows that both methods fail in different ways and that combining them reduces the error.
Kaplan and Ruback (1995) compared, across 51 highly leveraged transactions between 1983 and 1989, the value obtained by discounting cash flow forecasts with the price ultimately paid. DCFs using industry or market betas came within 15% of the price in 61–65% of cases. Multiples of comparable listed companies did so in only 37%, with an average downward bias of 17%. Their most cited conclusion: combining DCF and comparables explains more of the variance in prices than either method alone.
Liu, Nissim and Thomas (2002) showed that not all multiples are equal: forecast earnings dominate, sales are the worst value driver and choosing peers from the same industry improves every result.
Aswath Damodaran (NYU Stern) draws the fundamental distinction between valuing and pricing. In his words, “every multiple implicitly contains all the variables that drive a discounted cash flow valuation”, and “it is impossible to compare companies properly on a multiple unless we understand the nature of that relationship”. His recommendation is pragmatic: if the assignment is to price, then price; if it is to value, do not let market mood seep into the valuation. The worst outcome, he adds, is the hybrid that “is neither price nor value”.
Pablo Fernández (IESE) is harsher: multiples “produce highly dispersed and questionable results” and are only useful as a complement, to compare companies with similar cash flow valuations. He also notes that value always depends on expectations, so it may differ for each buyer.
Koller, Goedhart and Wessels (McKinsey, Valuation, 8th ed., 2025) turn the identity shown above into a selection criterion: a company is a true comparable if it has a similar return on capital and growth, not because it shares the same industry code.
The synthesis is clear. The DCF is more accurate when there are reliable forecasts and risk is calibrated with market data. Multiples are more accurate when there are genuine comparables and positive, stable earnings. And both perform better when used together.
When to use each method in an M&A process
The choice depends on the stage of the process, the information available and the purpose of the valuation.
|
Situation |
Primary method |
Role of the other method |
Why |
|---|---|---|---|
|
Initial screening of a target |
Trading and transaction multiples |
Simplified DCF to check whether the multiple implies a credible return |
There is no business plan or data access yet |
|
Non-binding offer |
Multiples on normalised EBITDA |
DCF on the seller’s plan, with scenarios |
A range must be set without full due diligence |
|
Binding offer after due diligence |
DCF on the plan as adjusted by due diligence |
Multiples as a market cross-check |
Due diligence adjustments (non-recurring items, normalised working capital, debt) are only properly captured in cash flows |
|
Mature sectors with many comparables |
Forward earnings multiples |
DCF as a consistency check |
Empirical evidence favours multiples when genuine comparables exist |
|
Cyclical companies |
DCF with mid-cycle earnings |
Multiples on average through-the-cycle earnings |
At the peak of the cycle the P/E looks low and the company looks cheap |
|
High growth or loss-making |
Probability-weighted scenario DCF |
Sales multiples for reference only |
Earnings multiples do not exist and sales multiples have the largest error |
|
Distressed companies |
Restructuring DCF and liquidation value |
Multiples with caution |
Healthy peers do not reflect default risk |
|
Impairment (IAS 36) |
DCF: value in use requires it |
Multiples for fair value less costs of disposal (IFRS 13) |
The standard defines the method |
|
Non-cash contributions and independent expert reports (art. 67 of the Spanish Companies Act, LSC) |
Both, reconciled |
— |
The expert must justify the value to third parties |
A rule of thumb: the more specific the company (synergies, restructuring plan, unique contracts), the more weight the DCF should carry. The more it resembles its peers and the more liquid the transaction market, the more reliable multiples become.
How to combine them without averaging
The practice of weighting the DCF at 50% and multiples at 50% is convenient and indefensible: it averages two errors without explaining either. Combining them properly means using each method to interrogate the other.
1. The DCF’s implied multiple. Divide the DCF value by next year’s EBIT and compare it with the peers. If the DCF implies 14x and the sector trades at 9x, one of the three variables in the formula (RONIC, g, WACC) is very different. It must be identified and justified, or corrected.
2. The reverse DCF. The opposite route is more revealing. Starting from the multiple the seller is asking for, M, solve the formula for the perpetual growth that price requires:
M(\mathrm{WACC}-g) = (1-t)\left(1-\frac{g}{\mathrm{RONIC}}\right) \;\Rightarrow\; g^{*} = \frac{M\cdot \mathrm{WACC}-(1-t)}{M-\frac{1-t}{\mathrm{RONIC}}}
If the seller asks for 12x EBIT, with a WACC of 9% and t of 25%, the price is only justified by perpetual nominal growth of 4.7% if the marginal return is 15%, or 3.5% if it is 30%. The negotiation stops being an argument about “the market multiple” and becomes a specific question: does the buyer believe this company will grow faster than the economy forever, earning 30% on every euro reinvested?
3. Ranges, not points. The result is presented as a football field chart by method: the DCF with its sensitivity to WACC and g, trading comparables with their interquartile range, and precedent transactions. The overlap between ranges is the negotiating zone. A discrepancy between them is not resolved by averaging: it is the most valuable information in the valuation.
4. The same bridge to equity value. Enterprise value less net financial debt, lease liabilities if the multiple is post-IFRS 16, debt-like provisions, normalised working capital adjustment and minority interests. Many differences attributed to the method are, in fact, differences in this bridge.
The DCF answers how much the company is worth if this plan is delivered. Multiples answer how much the market is paying today for something similar. A good valuation report needs both answers and, above all, an explanation of why they differ.
