DE GRUYTER Journal of Business Valuation and Economic Loss Analysis.

2020; 20190020

Andreas Schueler1

Valuation with Multiples: A Conceptual Analysis

1 Department of Economics and Management, Bundeswehr University Munich, Werner-Heisenberg-Weg 39, 85579 Neubiberg,

Germany, E-mail: [email protected]

Abstract:
Estimating the market price of a company with multiples is common practice. Especially if several multiples are
used simultaneously, the bandwidth of value estimates might be wide. The paper aims at narrowing down this
bandwidth with a conceptual analysis. I analyze the different ways to average peer multiples, the links between
common multiples (‘inter-multiple’ analysis), the relevance of their components (‘intra-multiple’ analysis) and
the resulting choice between a bottom-up and a top-down approach for deriving a multiple, and the impact of
differing capital structures.
Keywords: valuation, multiples, EV/Sales, EV/EBITDA, EV/EBIT, PE ratio, PB ratio
DOI: 10.1515/jbvela-2019-0020

1 Introduction
Multiples are used widely in practice for company valuation. Financial analysts, for instance, apply multiples
in addition to discounted cash flow (DCF) valuation to value a company and derive recommendations for in-
vestors. This is confirmed by empirical studies and surveys such as those by Asquith, Mikhail, and Au (2005),
Demirakos, Strong, and Walker (2004), and Imam, Barker, and Clubb (2008), or Mukhlynina and Nyborg (2016).
The paper seeks to analyze characteristics of valuations by multiples to enable analysts to check the consistency
of multiple-based valuations and narrow the range of share price estimates. Thus, the paper focuses on con-
ceptual issues and not, for instance, on empirical issues such as the forecast accuracy of different multiples, as
investigated by Alford (1992), Cheng and McNamara (2000), Lie and Lie (2002), and Liu, Nissim, and Thomas
(2002, 2007), and Chullen, Kaltenbrunner, and Schwetzler (2015), and many others.
The starting point of the paper is an alternative to averaging the multiples of the comparable companies to
a peer group multiple. The use of the harmonic mean is advocated based on its implications regarding invested
capital and return. This recommendation is also made, although in parts with different reasoning, by Baker
and Ruback (1999), Beatty, Riffe, and Thompson (1999), Liu, Nissim, and Thomas (2002), and Agrrawal et al.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

(2010). Second, several multiples are available for practitioners. The conceptual links between multiples are of
interest because they allow conclusions to be drawn on the consistency of valuation results, enable the analyst
to detect contradictions, and provide insights into how the valuation results depend upon the multiple used.
They can also help to understand why the empirical explanatory power and the coefficient of variation differ
between multiples as shown e. g. by Chullen, Kaltenbrunner, and Schwetzler (2015). Third, the paper addresses
the components of multiples and their reconciliation. It is shown that the valuation results depend on the
decision of the analyst to derive an average multiple either with a top-down approach or to aggregate it from
its components in a bottom-up approach. Finally, it is shown how multiples, which are not distorted by capital
structure, can be calculated following Holthausen and Zmijewski (2012). The results of the paper could be useful
not only to valuation practitioners but also to researchers because it provides, for instance, further insights into
the reasons why different multiples show variances in power to explain share prices.
The valuation by multiples is referred to as the market approach. It can be applied by using prices paid for
whole companies, or portions thereof that are considered controlling in nature in past transactions (completed
transaction method, CTM, or transaction multiples) or by using observable market capitalizations, which is
stock price times number of shares outstanding on a non-controlling basis (guideline public company method,
GPCM, or trading multiples). In the following, I will focus on the latter method.
Estimating share prices by applying the GPCM method (trading multiples) is a well-known process. In
the first step, comparable companies are identified, which should match with respect to risk, (especially for
equity multiples) capital structure, profitability, and solvency. Although the designation as a peer firm is not
necessarily industry dependent, such firms are often operating in the same industry. In the second step, the
multiple is calculated for each peer firm. Finally, the financial metric (e. g. free cash flow, EBIT, or earnings) of
the company to be valued is multiplied by the average multiple for the derivation of company value. Multiples
Andreas Schueler is the corresponding author.
© 2020 Walter de Gruyter GmbH, Berlin/Boston.

1
 Schueler DE GRUYTER

are defined as the ratio of company value divided by a financial metric such as sales, EBITDA, EBIT, or earnings.
Multiples can refer to total company value, i. e. the sum of the market value of equity plus debt from which cash
and marketable securities are sometimes subtracted (enterprise value, EV, or market value of invested capital,
MVIC), whereas equity value multiples use the market value of equity as the numerator. Common multiples
are EV/Sales, EV/EBITDA, EV/EBIT, PE ratio, and Price-to-Book (PB) ratio.
The paper aims at narrowing down the bandwidth of possible valuation results and addresses the steps of
the process to estimate share prices as follows: Commonly used multiples are introduced by using an example,
which is evolved throughout the paper. The next section focuses on the aggregation to a peer group multiple.
Then, the links between multiples will be analyzed. I will show that the results of a valuation based on compo-
nents of the multiple (bottom-up) differ from the results derived by the (entire) multiple (top-down). Finally, I
show unlevered multiples can be used as multiples free from distortions caused by capital structure effects.

2 Example
This paper focuses on the following multiples: EV to sales, EV to EBITDA, EV (or MVIC) to EBIT, PE ratio, and
PB ratio. Company A is to be valued by Guideline Public Company Method (trading multiples). I assume to
have identified three comparable firms, B, C, and D. Furthermore, I assume that the fiscal year of all companies
ended just recently.1 The value of debt and equity are as of the end of the fiscal year. The other financial data,
such as earnings, are defined as the consensus expectations for the current year, which has just begun (forward
looking multiples). However, the discussion in the remainder of the paper does not depend upon whether
trailing or forward-looking multiples are being used. I assume non-negative surpluses and zero growth. Thus,
the change (variation) in net working capital, capital expenditures exceeding depreciation, and sales growth
are not considered for simplicity for most of the paper. Table 1 provides the data, including a few standard
financial ratios like margins and rate of returns.

Table 1: Data in USD million.

Company Comparable companies Unweighted
A average
B C D
Market value of equity 82.0 80.0 24.0 62.0
Debt 3.0 30.0 18.0 24.0 24.0
Enterprise value (EV) 112.0 98.0 48.0 86.0
Leverage ratio (L) = debt divided by EV 26.8% 18.4% 50.0% 31.72%
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

1-L 73.2% 81.6% 50.0% 68.28%

Book value of equity 24.0 46.0 42.0 20.0 36.0
Cost of debt 4.5% 6.0% 5.0% 6.0% 5.7%
Levered cost of equity (rL ) 10.0% 9.0% 14.0% 11.0%
WACC 8.53% 7.94% 9.10% 8.5%
Sales 50.0 100.0 150.0 80.0 110.0
EBITDA 10.0 21.0 14.0 10.0 15.0
EBITDA margin (m) as percentage of sales 20.0% 21.00% 9.33% 12.50% 14.28%
Depreciation 3.0 8.0 5.0 3.0 5.3
Depreciation factor (d) as percentage of EBITDA 30.00% 38.10% 35.71% 30.00% 34.60%
EBIT 7.0 13.0 9.0 7.0 9.67
EBIT margin as percentage of sales 14.0% 13.00% 6.00% 8.75% 9.25%
Interest 0.14 1.80 0.90 1.44 1.38
Interest as percentage of EBIT (z) 1.93% 13.85% 10.00% 20.57% 14.81%
EBT 6.87 11.20 8.10 5.56 8.29
Tax rate (τC ) 30.0% 25.0% 35.0% 30.0% 30.00%
Taxes 2.06 2.80 2.84 1.67 2.43
Earnings 4.81 8.40 5.27 3.89 5.85
ROE = earnings divided by book value of equity 20.0% 18.3% 12.5% 19.5% 16.75%
Earnings before interest after adjusted taxes (EBIaT) 4.90 9.75 5.85 4.90 683.33%
ROA = EBIaT divided by equity plus debt 18.1% 12.8% 9.8% 11.1% 11.24%

To account for different tax regimes, differing corporate tax rates are employed, as are interest rates differing
between companies. Based on data from the peer companies, multiples can be calculated using the results as

2
 DE GRUYTER Schueler

provided in Table 2. For the EV multiples, debt employed of company A (USD 3 million) must be deducted to
move from entity value to the market value of equity.

Table 2: Value estimates based on multiples of company B, C, and D.

Multiples derived by the data of the comparable Equity value estimate in USD million for
companies company A based on multiples for B, C, and D
B C D B C D
EV/Sales 1.120 0.653 0.600 53.00 29.67 27.00
EV/EBITDA5.333 7.000 4.800 50.33 67.00 45.00
EV/EBIT 8.615 10.889 6.857 57.31 73.22 45.00
PE 9.762 15.195 6.166 46.91 73.02 29.63
PB 1.783 1.905 1.200 42.78 45.71 28.80

Although it may appear captious, it might seem implausible to classify a company as comparable if the range
of multiples varies considerably. Nevertheless, this is the case for the example, as Figure 1 shows.

Figure 1: Range of estimated values of equity in USD million employing multiples of companies B, C, and D.

Based on the multiples of the comparable companies, the range of equity values can be calculated. The equity
value varies from USD 27 million to USD 73 million . The lowest values are shown for the multiples of company
D, and the highest are found for company C. In my example, the market assigns the highest multiples to C
except for the sales multiple, despite a currently lower level of profitability in terms of current margins and rate
of returns (Table 1). This might be caused by expectations about a higher level of profitability in the future.
Because the large variation is inadequate, we need to narrow the range. The first step on that path is the
aggregation to an average value.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

3 Aggregation to a Peer Group Average

Table 4 shows company-specific multiples and average values. Alongside the weighted and arithmetic (un-
weighted) averages, the median and harmonic mean are presented. The median is less prone to outliers. How-
ever, for a small sample, such as mine, the median is not helpful. It does not make sense to drop two of the
three observations because they are not the median value. Table 4 also contains the harmonic mean, which is
the reciprocal of the arithmetic mean.
The harmonic mean has some useful properties: Multiplying net income of the company being valued by
the PE ratio is equivalent to dividing net income by the reciprocal of the PE ratio, of course. Using the formula
for valuing a perpetuity (eq. (1)), also known as the Gordon growth formula of the Dividend Discount Model
(DDM) if we assume that earnings equal dividends, it becomes apparent that the reciprocal of the multiple
equals the discount rate for capitalizing these earnings minus the growth rate.

Earnings
V= (1)
Discount rate − growth rate

Therefore, valuing a company by multiples is comparable to a capitalization of earnings assuming that earnings
remain constant (or increases by a constant growth rate) and that the average difference between discount rate
and growth rate for the comparable companies matches the difference for the company to be valued.
Coming back to aggregating the PE ratios of the peers, the aggregation should focus on the reciprocal of the
PE ratio, which represents the denominator of (1). As Table 3 and Table 4 illustrate, there is a difference between

3
 Schueler DE GRUYTER

the average PE ratio and the average inverse PE ratio. Before moving on, it should be clarified why this is the
case:
The function of the reciprocals of the PE ratios (PE−1 ) is convex. The secant line between two points of
a convex function lies above the function itself. The arithmetic average lies on the secant line. Therefore, the
arithmetic average of the PE ratios is always higher than the average value based on reciprocals that lies on the
function itself. This relationship and the conclusion of interest here ‘the average of the reciprocals exceeds the
reciprocal of the average’ is known as Jensen’s inequality (see, for example, Agrrawal et al. 2010).
In line with Jensen’s inequality, the average of the inverse PE ratios of companies B, C, and D is 0.1101. The
inverse of 0.1101 is 9.079, which is the harmonic mean of the PE ratios. It is lower than the arithmetic average
of the PE ratios, 10.374.

Unweighted averagePE−1 = 31 ( M 1 + M 1 + M1 )
PE,B PE,C PE,D
= 31 ( 9.762
1 1
+ 15.195 1
+ 6.166 ) = 0.1101; (2)
−1
HMPE = 0.1101 = 9.079.

Figure 2 illustrates Jensen’s inequality and its implications for averaging the PE ratios of companies B, C, and
D (see Schwetzler (2003) for a similar illustration). The convex function relates inverse PE values (x-axis), using
the harmonic mean, i. e. the average denominator of (1), to the price for one USD of earnings (y-axis). Thus, the
average value for one USD of earnings should be 9.079. The dotted line above the convex function would lead
to the arithmetic average of 10.374. This is too high, because the convex value function only justifies a value, i. e.
an average PE ratio, of 9.079.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Figure 2: Illustrating Jensens’s inequality and the implications for the average PE ratio.

In light of the link to the capitalization of earnings, which uses the average cost of equity minus the growth rate,
the harmonic mean should be used. In addition, the use of the harmonic mean can be justified empirically as in
Liu, Nissim, and Thomas (2002), although it should be noted that empirically, the harmonic mean is sensitive
to negative outliers. Beatty, Riffe, and Thompson (1999) point out that the harmonic mean should lead to better
value predictions as the measurement error in earnings is shown in the denominator of the PE ratios leading
to upward biased predictions. The inverse PE Ratio avoids this problem.
Finally, the use of the harmonic mean can be justified by looking at the surplus and investment implied, as
shown by, for instance, Pratt, Reilly, and Schweihs (2000, 244), or Agrrawal et al. (2010):
An investor investing USD 1 million in each of the companies B, C, and D is entitled to the following net
income (NI):

4
 DE GRUYTER Schueler

Investment
in USD million in B
1
⏞ 1 1
82
∗ 8.4
⏟+ ∗ 5.27 + ∗ 3.89 = 0.3304 (3)
⏟
No. of shares in million NI B 80
⏟⏟⏟⏟⏟ ⏟ 24
⏟⏟⏟⏟
multiplied with ...C... ...D...
share price B

The investor receives a share of net income according to his stake in the market value of equity. Market capital-
ization and net income are derived from Table 1. Total net income attributable to the investors (0.3304) divided
by his investment (3) equals the expected rate of return of (0.3304/3) = 0.1101. This is the reciprocal of the har-
monic mean of the PE ratio: 0.1101−1  = 9.079.
However, if one presumes that the investor plans to achieve a net income of USD 1 million from each invest-
ment, USD 31.112 million need to be invested:

1 1 1
∗ 82 + ∗ 80 + ∗ 24 = 31.112 (4)
8.4
⏟ 5.27
⏟⏟⏟⏟⏟ ⏟ 3.89
⏟⏟⏟⏟
Investment in B in USD million: ...C...∶ ...D...∶
9.762 15.195 6.166

Now, the expected rate of return is USD 3 million/USD 31.112 million = 0.0964. The inverse of this rate of return,
0.0964−1  = 10.37, is equal to the arithmetic average of the PE ratios of companies B, C, and D. The link between
(eq. (3)) and the usual way to calculate the arithmetic mean is:

⎛
⎜ 82 80 24 ⎞ ⎟ 1 1
⎜
⎜ + + ⎟
⎟ = 31.112 ∗ = 10.37 (5)
⎜ 8.4
⏟ 5.27
⏟ ⏟ ⎟3
3.89 3
⎝PE B 9.762 PE C 15.195 PE D 6.166⎠

We can conclude that by using different concepts for averaging PE ratio, one implicitly assumes different levels
of investment. By using the harmonic mean, one assumes that the same amount is invested in each of the
comparable companies. Using the arithmetic mean implies an investment necessary to receive the same amount
of earnings from each share of B, C, and D. Because the first assumption is more plausible, the harmonic mean
should be used instead of the arithmetic mean.
As Table 3 shows, due to Jensen’s inequality the use of arithmetic (unweighted) average will overvalue the
valuation subject, Company A. The arithmetic average of each multiple is higher than the harmonic mean of
the multiples. As already pointed out above, the arithmetic average leads to upward biased results.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

Table 3: Company specific and average multiples.

Comparable companies Peer Group
B C D Unweighted Weighted Median Harmonic
average average mean
EV/Sales 1.120 0.653 0.600 0.791 0.782 0.653 0.733
EV/EBITDA5.333 7.000 4.800 5.711 5.733 5.333 5.569
EV/EBIT 8.615 10.889 6.857 8.787 8.897 8.615 8.481
PE 9.762 15.195 6.166 10.374 10.594 9.762 9.079
PB 1.783 1.905 1.200 1.629 1.722 1.783 1.563

That the choice of the averaging approach is relevant is confirmed by the summary in Table 4.

Table 4: Valuation estimates for different aggregation methods in USD million.

Unweighted Weighted Median Harmonic
average average mean
Enterprise Value A
EV/Sales 39.56 39.09 32.67 36.67
EV/EBITDA 57.11 57.33 53.33 55.69
EV/EBIT 61.51 62.28 60.31 59.37

5
 Schueler DE GRUYTER

PE 52.85 53.91 49.91 46.63

PB 42.10 44.33 45.78 40.51

Equity Value A
EV/Sales 36.56 36.09 29.67 33.67
EV/EBITDA 54.11 54.33 50.33 52.69
EV/EBIT 58.51 59.28 57.31 56.37
PE 49.85 50.91 46.91 43.63
PB 39.10 41.33 42.78 37.51

Figure 3 shows the corresponding bandwidth of equity values.

Figure 3: Range of estimated values of equity in USD million employing different methods for aggregating peer group
multiples (minimum and maximum values of the gray bars: derived from Table 2; black bars: from Table 4. HM stands for
harmonic mean; MD: median; UA: unweighted average; WA: weighted average).

It is evident that my valuation still does not lead to a narrow range of values. The gray bars in Figure 3 show
the range of values using the non-aggregated multiples of comparable companies. The black highlighted areas
represent the value range based on the aggregated multiples. They do not overlap in my example. My prelimi-
nary results do not provide a concise basis for estimating the maximum price that a buyer would be inclined to
pay or the minimum price a seller would accept. Even if we used only the harmonic mean, the valuation results
depend upon the multiple used.

4 Links between Different Multiples

In practice, not all multiples are used at the same time. While selecting the appropriate multiple(s), industry
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

specifics, as well as the empirical explanatory power of different multiples, and their empirical fit shown, for
example, by their coefficient of variation are to be considered. However, a prerequisite for analyzing the dif-
ferences in conceptual and empirical fit it is helpful to develop an understanding about the links between the
multiples available for the valuation at hand.
It is known that multiples can be reconciled (see, for instance, Koller, Goedhart, and Wessels 2015, 327;
Soffer and Soffer 2003, 423–427). Nevertheless, to the best of my knowledge no contribution to the literature
discusses the links between all five multiples in detail. An understanding of the linkages is essential for laying
the groundwork for verifying consistency among the multiples applied. The basic idea can be illustrated for
the transition from EV/Sales to EV/EBITDA multiple. After multiplying with the EBITDA multiple and its
reciprocal, it becomes evident that the EV/Sales multiple equals the product of EBITDA margin and EBITDA
multiple:

EV 
EV EBITDA EV EBITDA EV
= ∗  ∗ = ∗ (6)
Sales Sales 
EV EBITDA Sales
⏟⏟⏟⏟⏟ EBITDA
EBITDA−margin

Furthermore, it is interesting for practical usage to determine how the succeeding multiple can be deduced
from the preceding one. For the transition from the EBITDA multiple back to the sales multiple, the EBITDA
multiple must be extended analogously by multiplying with the reciprocal of the sales multiple:

EV 
EV Sales EV Sales EV
= ∗  ∗ = ∗ (7)
EBITDA EBITDA 
EV Sales EBITDA
⏟⏟⏟⏟⏟ Sales
reciprocal of
EBITDA − margin (m−1 )

6
 DE GRUYTER Schueler

The EV/EBITDA multiple therefore equals the product of the reciprocal of the EBITDA margin (m−1 ) and
EV/Sales multiple. This idea can be employed for other multiples as well:

– The EV/EBIT multiple is equal to the EV/EBITDA multiple times the reciprocal of (1-d). The depreciation
factor d is defined as depreciation in percent of EBITDA.

EV 
EV EBITDA EV ( (((
EBITDA EV
EBIT
=
EBIT
∗

EV
∗
EBITDA
=
( ( (( (1 − d) ∗ EBITDA
EBITDA
(8)
⏟⏟⏟⏟⏟⏟⏟⏟⏟
reciprocal of
′
1−depreciation f actor ′

Analogously the EV/EBITDA multiple can be written as the product of (1-d) and EV/EBIT. The relation can be
broken down to the EV/Sales multiple:

EV −1 EV −1 EV
= (1 − d) ∗ = m−1 ∗ (1 − d) ∗ (9)
EBIT EBITDA Sales
– The transition of the EV/EBIT multiple to the PE ratio is more complex because on the one hand, interest
payments and taxes must be considered, and on the other hand, the altered capital base – equity value
instead of entity value – must be accounted for. With the interest factor z, the equity-to-EV ratio in terms of
market values (P/EV) and the leverage ratio L, we obtain

P P
= Earnings ∗ EBIT EV = EBIT ∗ P ∗ EV
∗ EBIT
Earnings EV Earnings EV EBIT
EBIT
= EBIT−Interest−Tax P ∗ EV
∗ EV EBIT (10)
−1
= (1 − z)−1 ∗ (1 − 𝜏C ) ∗ EV−F EV
EV
∗ EBIT
−1
= (1 − z)−1 ∗ (1 − 𝜏C ) ∗ (1 − L) ∗ EBIT EV

The PE ratio can be traced back to the ratio of EV/Sales as well, substituting the EBIT multiple by eq. (8):

−1
P
Earnings
= (1 − z)−1 ∗ (1 − 𝜏C ) EV
∗ (1 − L) ∗ EBIT
−1 −1 −1 (11)
=m ∗ (1 − d) ∗ (1 − z)−1 ∗ (1 − 𝜏C ) EV
∗ (1 − L) ∗ Sales

– The transition of the PE ratio to the PB ratio is less demanding because the latter equals the product of PE
ratio and ROE:
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

P = P
∗
Earnings
∗
P
=
Earnings
∗
P
(12)
BEq BEq 
P Earnings ⏟⏟⏟BEq⏟⏟ Earnings
ROE

Knowing the linkages between different multiples is useful because it highlights the components of a multiple.
If the EV/Sales, EV/EBITDA, and EV/EBIT multiple are used simultaneously, the implied EBITDA margin,
EBIT margin, or expense ratios can be compared to historical observations or to competitors’ data. Further-
more, based on a preceding multiple, such as EV/Sales, succeeding multiples can be determined based on
margins. Overall, the possibilities for checking the transparency and plausibility of a valuation using multiples
are increased.

5 Top-down Vs. Bottom-up Derivation of Multiples

Unfortunately, for the analyst using multiples, a multiple depends upon the way it is calculated. Table 5 shows
that it matters whether a multiple is calculated in one-step (top-down) or in several steps (bottom-up) by starting
from other multiples based upon the links discussed above. If the EV/EBITDA multiple is derived following
eq. (7) by multiplying the inverse of the EBITDA margin (unweighted average 7.825) with the average EV/Sales
multiple (0.791, see Table 3), we obtain a multiple of 6.191 and an equity value of USD 58.91 million instead of a
multiple of 5.711 and an equity value of USD 54.11 million. The equity value follows from subtracting net debt
from enterprise value.

Table 5: Multiples derived directly or starting from other multiples (bottom-up approach).

7
 Schueler DE GRUYTER

Unweighted Enterprise Equity value

average value A in A in USD
USD million million
EV/EBITDA (one step) 5.711 57.11 54.11
to EV/EBITDA from EV/Sales
m−1 * EV/Sales 6.191 61.91 58.91
EV/EBIT (one step) 8.787 61.51 58.51
to EV/EBIT from EV/Sales
m−1 * (1-d)−1 * EV/Sales 9.491 66.44 63.44
[m * (1-d)]−1 * EV/Sales 9.437 66.06 63.06
… from EV/EBITDA
(1-d)−1 * EV/EBITDA 8.756 61.29 58.29
PE ratio (one step) 10.374 52.85 49.85
to PE ratio from EV/Sales
m−1 * (1-d)−1 * (1-z)−1 * (1- τC )−1 * (1-L) * EV/Sales 10.934 55.54 52.54
… from EV/EBITDA
(1-d)−1 * (1-z)−1 * (1- τC )−1 * (1-L) * EV/EBITDA 10.087 51.47 48.47
… from EV/EBIT
(1-z)−1 * (1- τC )−1 * (1-L) * EV/EBIT 10.123 51.64 48.64
PB ratio (one step) 1.629 42.10 39.10
to PB ratio from EV/Sales
m−1 * (1-d)−1 * (1-z)−1 * (1- τC )−1 * (1-L) * ROE * EV/Sales 1.832 46.96 43.96
… from EV/EBITDA
(1-d)−1 * (1-z)−1 * (1- τC )−1 * (1-L) * ROE * EV/EBITDA 1.690 43.55 40.55
… from EV/EBIT
(1-z)−1 * (1- τC )−1 * (1-L) * ROE * EV/EBIT 1.696 43.70 40.70
… from PE ratio
ROE * PE ratio 1.738 44.71 41.71

As a result, new valuation estimates emerge and prolong the list of possible results. The PE multiple, for ex-
ample, can be derived directly by using market capitalization and net income of the peer group companies B,
C, and D or by employing the modular approach starting with the preceding multiple and deriving the PE
multiple employing margins and expense ratios.
What is the reason for the deviations? The definition of the covariance answers that question: The covariance
of two variables X and Y equals the difference between expected value of Y times the expected value of X and
the expected value of Y times X:
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

𝜎XY = E [XY] − E [X] ∗ E [Y] (13)

E [XY] = E [X] ∗ E [Y] + 𝜎XY (14)

Eq. (14) results by applying eq. (13) to the relation between the EV/Sales multiple and EV/EBITDA multiple
as in eq. (6). The covariance between EBITDA margin and EV/EBITDA multiple equals −0.024.

E [EV/Sales] = E [m] ∗ E [EV/EBITDA] + 𝜎m;EV/EBITDA

= 0.1428
⏟⏟⏟⏟⏟ ∗ 5.711
⏟⏟ + (−0.024) (15)
0.815
= 0.791

The average EBITDA margin (0.1428, see Table 1) multiplied by the average EV/EBITDA multiple (5.711, see
Table 5) plus the covariance (−0.024) confirms the average EV/Sales multiple (0.791, see Table 3). A covariance
different from zero signals that the components of the EV/Sales multiple of the comparable companies are not
independent of each other. The analogous transition from EV/Sales multiple to EV/EBITDA multiple is

E [EV/EBITDA] = E [m−1 ] ∗ E [EV/Sales] + 𝜎m−1 ;EV/Sales

−1
= 0.1278
⏟⏟⏟⏟⏟⏟⏟ ∗ 0.791 + (−0.480) (16)
6.191
= 5.711

8
 DE GRUYTER Schueler

6 Preliminary Results
The stepwise process of the bottom-up approach increases the transparency and verifiability of the valuation.
It clarifies that comparability does not end, but rather starts with comparability in terms of the surplus used. It
also makes clear that different combinations of components might exist that lead to the same or similar multiple
values.
A technical result derived above is that top-down and bottom-up approach result in identical multiples
producing identical valuation results only if the components are independent of each other (covariance: 0).
In addition, the granularity of the stepwise bottom-up approach, i. e. the number and length of the steps, is
value relevant. Although the links illustrated by eq. (6) to (12) work in case of a company on a stand-alone
basis, they do not provide unanimous valuation results for the peer group multiples, as the discussion in the
preceding section has shown. Figure 4 shows for the data of the example, that the value range defined by
different valuation approaches (black bars) is wider than it was in Figure 3.

Figure 4: Range of estimated values of equity in USD million using different averages, and top-down and bottom-up
approach (minimum and maximum values of the gray bars: derived from Table 2; black bars: Table 4 and Table 5. HM
stands for harmonic mean; MD: median; UA: unweighted average; WA: weighted average).

The interval of aggregated values (black bars) depends on the definition of the average value, which was also
true for the results shown in Figure 3. In addition, the level of aggregation matters. In the following, I will only
use the harmonic mean due to its superiority.
The position of the value intervals is determined by the deviation of the components implied by the multiple
for company A from those for the peer group. The following observations hold for the example:

– Equity values based on the average EV/EBITDA multiples (black bars) surpass the values derived with the
EV/Sales multiple because the EBITDA margin of company A (20%) exceeds the average margin of the peer
group companies (unweighted average: 14.28%). Table 4 confirms this observation. Because the EV/Sales
multiple does not account for the higher profitability of company A, the value measured by the EV/EBITDA
multiple provides a more reliable result. In line with eq. (6), this could be illustrated by multiplying the peer
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

group EV/EBITDA ratio, for example, the unweighted mean of 5.711, by the EBITDA margin of A (20%).
One receives an adjusted EV/Sales multiple for A of 1.142; after multiplication by the sales of A, (50) results
in an enterprise value A of USD 57.11 million (equity value of USD 54.11 million), which is equal to the
result delivered by the EBITDA multiple. Thus, the adjusted bottom-up sales multiple fits to the EBITDA
margin of company A.
– The EV/EBIT based interval shows higher company values than the EV/EBITDA based interval because the
depreciation factor of A (30%) is smaller than that of the peer group (the unweighted average equals 34.6%).
Therefore, the reciprocal of (1-d), which is necessary for linking EV/EBITDA and EV/EBIT, is smaller. That
the use of the EV/EBIT multiple leads to a higher entity value of A than the use of the EV/EBITDA multiple
(see Table 4) can also be shown by rearranging eq. (7) and employing the firm specific depreciation factor
of A:

EV EV
= (1 − dA ) = (1 − 0.3) 8.787 = 6.151
EBITDAA EBITPeer Group

The EBITDA of A (10) multiplied by the adjusted EV/EBITDA multiple of 6.151 results in an entity value of A
of USD 61.51 million, such as the EV/EBIT multiple (Table 4).

– For the transition of the EV/EBIT multiple to the PE ratio (eq. (9)) interest payments, taxes and leverage ratio
are to be considered. Interest in percentage of EBIT is smaller for company A (2%) as it is for the peer group
(unweighted mean: 14.81%). Therefore, the reciprocal of (1-z) is smaller. All else being equal, this would
result in a PE-based sub-interval right to the EV/EBIT interval because the earnings of A are less reduced
by interest expenses and are therefore higher than the average PE ratio implies. The tax rate does not cause

9
 Schueler DE GRUYTER

value differences here because it is the same for company A and the peer group tax rate (30%). Unlike the tax
rate, the leverage ratio cannot be assumed to be identical. Accounting for the difference in the leverage ratio is
more challenging. The leverage ratio is 31.72% for the peer group, but the ratio for company A is not known
from the beginning, as the entity value and the value of equity are not known, but the valuation results
are. A circularity problem emerges. In light of the valuation results for A based on EV/EBIT multiples, as
stated in Table 4, one can conclude that the leverage ratio of A must be considerably smaller than 31.72%
because the value of debt is very small in relation to the entity value. All else being equal, this effect results
in a shift of the estimated values to the left, as the value increasing effect of the lower interest component
is dominated by the value decreasing effect of the lower leverage ratio. The range of equity values based
on the PE ratio is smaller than the value range according to the EV/EBIT multiple. Stated differently, as the
leverage ratio of the peer group surpasses the leverage ratio of A, too much debt is assigned to company A
by using the PE ratio. This distortion can be avoided if multiples based upon total company value (EV or
MVIC multiples) are used.

Thus far, in accordance with common practice, I have neglected the impact of tax shields caused by debt fi-
nancing. In the last section, I address that problem.

– The values estimated by the PB ratio are smaller than the values based on the PE ratio, as the return on equity
of A (20%) exceeds the ROE of the peer group (unweighted mean: 16.75%). This is the consequence of the
rather low book value of equity of company A, which, after being multiplied by the rather low peer PB ratio,
leads to a low estimate of equity value. The lower book value of company A could be caused, for example, by
intangible assets if their value is not fully captured by the book value. Put differently, the lower peer group
ROE leads to a PB ratio that is too low for A. An adjusted PB ratio that avoids this problem can be written
according to eq. (12) based on the unweighted mean of the PE ratio of 10.374:

P P
= ROEA ∗ = 0.2002 ∗ 10.3744 = 2.077
BEq E

The book value of equity of A (24) multiplied by the adjusted PB ratio of 2.077 results in a market value of equity
of USD 49.85 million, like the PE ratio does (Table 4).
Value estimates that are built upon more specific multiples, such as EBIT multiples, reflect the characteristics
of the company being valued rather than simpler multiples such as sales multiples. Splitting up a multiple into
its components can explain differing valuation results and enables the derivation of reconciled multiples, which
can be adjusted to the specifics of the company to be valued. Therefore, a sales multiple adjusted to the margin
of the company to be valued or an EBITDA multiple adjusted to the depreciation factor can be derived. The
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

PE ratio is distorted if the capital structure of the company to be valued differs significantly from the capital
structure of its peers. We address that problem in the last section. The PB ratio is distorted by differences in
profitability, as measured by ROE. Avoiding that problem requires the use of an adjusted PB ratio. Because this
only confirms the results obtained by applying the PE ratio, the use of the adjusted PB ratio is inefficient and
unnecessary. The sales multiple adjusted for differing margins and the EBITDA multiple adjusted for differing
depreciation factors also suffer from that weakness.
In summary, the value of the equity of company A still ranges from USD 33.7 million to USD 56.4 million
(Figure 5). Differences in value occur due to alternative approaches in determining the average peer group
multiple, and due to the choice between the top-down and the bottom-up approach. It should be noted that the
range of values would be even broader if all possible variations have been considered. Examples for additional
alternatives could be developed based on the direction of the transition (e. g. not only from the EV/Sales multi-
ple to the EV/EBITDA multiple but also from the EV/EBITDA multiple to the EV/Sales multiple) and the use of
other measures for the aggregation of the components (weighted mean, median, harmonic mean). Fortunately,
the resulting broad range of equity values can be narrowed. As stated above, the sales multiple does not account
for the above-average margin of company A, but the EBITDA multiple does. The EBIT multiple accounts for
the different deprecation ratio additionally. The PE ratio is distorted by the difference in leverage between the
peer group and company A. Because the PB ratio builds upon ROE, it is also affected by the leverage effect and,
thus, by differing leverage ratios. The difference between ROE and the cost of equity in the numerator of that
equation may be appealing economically. However, it is distorted by differences in profitability between the
company to be valued and its peers.

10
 DE GRUYTER Schueler

Figure 5: Range of estimated values of equity in USD million (minimum and maximum values of the gray bars: derived
from Table 2; highlighted values: harmonic means (HM) according to Table 4, Table 5).

7 Dealing with Differing Capital Structures

Valuing equity by the Adjusted-Present-Value (APV) method, a variation of the DCF approach, requires esti-
mating the enterprise value of the unlevered company (EVU ) first. The unlevered enterprise value is adjusted
to reflect the influence of debt financing by adding the present value of tax shields (VTS ). This sum equals the
enterprise value and reflects debt financing (EV). After subtracting debt, we obtain the value of equity. To use
that idea for a multiple-based valuation following Holthausen and Zmijewski (2012), the value of the unlevered
enterprise must be calculated for all peer companies. The estimated present value of the tax shields must be
subtracted from the sum of the market capitalization and debt for that purpose. Because a multiple refers to a
perpetuity, periodic tax shields can also be considered constant. Thus, the value of the tax shields is simply the
tax rate multiplied by debt employed. The unlevered enterprise value of company B, for example, is

EV = Market value of equity + debt

= USD 82 million + USD 30 million = USD 112 million
(17)
EVU = EV − VTS = EV − 𝜏 ∗ D
= USD 112 million − 0.25 ∗ USD 30 million = USD 104.5 million

EVU /EBITDA and EVU /EBIT multiples are defined analogously. The unlevered version of the PE ratio is the
ratio of EVU to unlevered net income (EBIaT), defined as EBIT multiplied by the factor (1-τ). I refer to all mul-
tiples based upon unlevered enterprise values as unlevered multiples. These multiples can be applied to the
surplus of the company being valued (EBITDA, EBIT, or EBIaT), resulting in the unlevered EV of company A.
Adding the estimated value of the tax shields of A and subtracting the debt of A leads to the value of equity of
A:
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

EVU /EBIaTHM = 10.823

EBIaTA = EBITA (1 − 𝜏A ) = USD 7 million (1 − 0.3) = USD 4.9 million
EVU,A = USD 4.9 million ∗ 10.823 = USD 53.03 million (18)
EVA = EVU,A + VTS,A = USD 53.03 million + 0.3 ∗ USD 3 million = USD 53.93 million
EA = EVA − DA = USD 53.93 million − USD 3 million = USD 50.93 million

Table 6 shows the results for the other multiples.

Table 6: Valuation estimates based on unlevered APV based multiples in USD million.
Multiples B C D Harmonic mean
EVU /EBITDA 4.976 6.550 4.080 5.011
EVU /EBIT 8.038 10.189 5.829 7.612
EVU /EBIaT 10.718 15.675 8.327 10.823

Value of equity for B C D Harmonic mean

company A

EVU /EBITDA 47.66 63.40 38.70 48.01

EVU /EBIT 54.17 69.22 38.70 51.18
EVU /EBIaT 50.42 74.71 38.70 50.93

11
 Schueler DE GRUYTER

The resulting bandwidths are shown by Figure 6.

Figure 6: The final range of values of equity in USD million based on unlevered multiples.

Although value estimates still vary for the unlevered multiples, the remaining deviation is small and can be
explained in a straightforward manner: The lower value estimates based on the EVU /EBITDA multiple are
caused by the lower depreciation rate of A in comparison to the peer group. The differing tax rates cause the
difference between the value range according to EVU /EBIT and EVU /EBIaT. In case of uniform deprecation
ratios and tax rates, the application of all three multiples result in the same range of values. The unlevered
enterprise value amounts to 10.823 times USD 4.9 million = USD 53.03 million and an equity value of USD
50.93 million.
The results are not distorted by differences in the capital structure between the company to be valued and
its peers.

8 Conclusion
Financial analysts regularly use multiples derived from comparable companies to value other companies in a
similar industry or line of business. The analyst should be prepared to check the plausibility and consistency of
a multiple-based valuation despite the remaining well-known challenges, such as identifying comparable com-
panies and the implied assumptions of a perpetuity. The paper shows that the aggregation of the multiples of
comparable companies should be based on the harmonic mean. This conclusion can be justified empirically
and conceptually. The harmonic mean avoids the upward bias of the arithmetic average and implies an equal
contribution of each peer company to the mean. Analyzing the relationships between multiples employed and
their reconciliation can increase transparency and serve as a means for evaluating the consistency of the con-
ceptual alternatives. These alternatives inherent in using multiples (the choice between different definitions
of the average peer group multiple, the choice between top-down and bottom-up approach, and the choice
between using levered or unlevered multiples) leads to a considerable bandwidth of equity values. Using a nu-
merical example, the paper demonstrates how to narrow down this bandwidth. Differences in capital structure
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

between the comparable companies and the company being valued can be avoided by using unlevered APV
based multiples.

Notes
1 For simplicity, I assume that the interest-bearing debt equals net debt because no liquid assets exist and no interest income is considered.
Therefore, EV and MVIC are equal to the sum of market value of equity and debt.

References
Agrrawal, Pankaj, Richard Borgman, John M. Clark, and Robert Strong. 2010. “Using the Price-to-Earnings Harmonic Mean to Improve Firm
Valuation Estimates.” Journal of Financial Education 36 (3–4): 98–110.
Alford, Andrew W. 1992. “The Effect of the Set of Comparable Firms on the Accuracy of the Price-earnings Valuation Method.” Journal of Ac-
counting Research 30 (1): 94–108.
Asquith, Paul, Michael B. Mikhail, and Andrea S. Au. 2005. “Information Content of Equity Analyst Reports.” Journal of Financial Economics 75
(2): 245–82.
Baker, Malcom, and Richard S. Ruback. 1999. “Estimating Industry Multiples.” Working Paper. Harvard University.
Beatty, Randolph P., Susan M. Riffe, and Rex Thompson. 1999. “The Method of Comparables and Tax Court Valuations of Private Firms: An
Empirical Investigation.” Accounting Horizons 13 (3): 177–99.
Cheng, Agnes C. S., and Ray McNamara. 2000. “The Valuation of Accuracy of the Price-Earnings and Price-Book Benchmark Valuation Meth-
ods.” Review of Quantitative Finance and Accounting 15 (4): 349–70.

12
 DE GRUYTER Schueler

Chullen, Alberto, Hannes Kaltenbrunner, and Bernhard Schwetzler. 2015. “Does Consistency Improve Accuracy in Multiple-based Valua-
tion?” Journal of Business Economics 85 (6): 635–62.
Demirakos, Efthimios G., Norman C. Strong, and Martin Walker. 2004. “What Valuation Models Do Analysts Use?” Accounting Horizons 18 (4):
221–40.
Holthausen, Robert W., and Mark E. Zmijewski. 2012. “Valuation with Market Multiples: How to Avoid Pitfalls When Identifying and Using
Comparable Companies.” Journal of Applied Corporate Finance 24 (3): 26–38.
Imam, Shahed, Richard Barker, and Colin Clubb. 2008. “The Use of Valuation Models by UK Investment Analysts.” European Accounting Re-
view 17 (3): 503–35.
Koller, Tim, Marc Goedhart, and David Wessels. 2015. Valuation: Measuring and Managing the Value of Companies, 6th ed. Hoboken: McKinsey &
Company.
Lie, Erik, and Heidi J. Lie. 2002. “Multiples Used to Estimate Corporate Value.” Financial Analysts Journal 58 (2): 44–56.
Liu, Jing, Doron Nissim, and Jacob Thomas. 2002. “Equity Valuation Using Multiples.” Journal of Accounting Research 40 (1): 135–72.
Liu, Jing, Doron Nissim, and Jacob Thomas. 2007. “Is Cash Flow King in Valuations?” Financial Analysts Journal 63 (2): 56–68.
Mukhlynina, Lilia, and Kjell G. Nyborg. 2016. “The Choice of Valuation Techniques in Practice: Education versus Profession.” Working Paper.
University of Zurich.
Pratt, Shannon P., Robert F. Reilly, and Robert P. Schweihs. 2000. Valuing a Business: The Analysis and Appraisal of Closely Held Companies, 4th ed.
New York et al: McGraw-Hill.
Schwetzler, B. 2003. “Probleme der Multiple-Bewertung.” FinanzBetrieb 5: 79–90.
Soffer, Leonard C., and Robin J. Soffer. 2003. Financial Statement Analysis: A Valuation Approach. Upper Saddle River, NJ: Prentice Hall.
Automatically generated rough PDF by ProofCheck from River Valley Technologies Ltd

13