Measuring Portfolio Valuation

David Bernstein
1786 Grape Street
Denver Colorado 80220

Email
Bernstein.book1958@[Link]

Phone
202 413 5492

The goals of this paper are to discuss issues pertaining to the valuation of portfolios, mutual
funds, and exchange traded funds (ETFs). The analysis presented here indicates that many
commonly reported ETF and mutual fund PE ratios used by investment companies are highly
misleading. The paper discusses correct and incorrect ways to measure the PE ratio of a
portfolio of stocks. Statistical tests based on groups of firm PE ratios are shown to be misleading
because PE ratios are undefined when earnings are negative and outliers when earnings are
small. Tests based on the ratio of the difference of market cap and earnings to market cap are
shown to be more informative than tests based on PE ratios.

David Bernstein retired as an economist from the U.S. Treasury in 2012. He is the publisher of
several blogs including [Link], [Link] and
[Link]. He and his family now live in Denver Colorado. Email
Bernstein.book1958@[Link] and Phone 202 413 5492.

Electroniccopy
Electronic copyavailable
available at:
at: [Link]
[Link]
 Section One: Introduction

Most reputable financial analysts acknowledge that market valuations at the writing of this
article in the fall of 2017 are high compared to past market valuations. There is, however, wide
disagreement, about the extent of current market valuations and the likelihood of a correction in
the near term.

One of the reasons for the disagreement is that analysts are free to use several types of valuation
measures. PE ratios, the most commonly used measure of a stock’s valuation, can be based on
trailing or forward earnings or cyclically adjusted earnings adjusted for inflation. It is natural that
different statistical measures provide a different view on the level of the market.

A second factor impacting the measurement of portfolio PE ratios and comparisons between the
valuations of groups of firms stems from the limitation of the PE.
The PE ratio of a firm has no economic meaning when the firm has negative earnings. Moreover,
the PE ratio of a firm will be an extreme outlier when earnings are near zero.

A third set of issues involves the methodology used to construct a PE index for a fund. Some
investment companies omit all information about firms with negative earnings when constructing
their fund PE ratio. Some firms appear to take a weighted average of PE firms in their fund and
often these firm specify an upper bound for firms with large PE ratios. PE ratios calculated by
major investment firms using these procedures are arbitrary and often misleading.i

This paper discusses issues related to the valuation of firms and portfolios. I consider the
correct way to evaluate the PE ratio of a portfolio and show that some commonly used
techniques provide misleading answers. I derive a new formula for a portfolio PE ration that is
consistent with the ratio of the sum of firm market caps to the sum of firm earnings. I
demonstrate that problems associated with hypothesis tests using data on PE ratios can be
resolved by basing tests on an alternative financial ratio – the ratio of the difference between
market cap and earnings to market cap.

Section Two: Issues related to individual firm PE ratios

The PE ratio is the most often cited measure of the valuation of a stock or a portfolio of stocks.
This ratio has no economic meaning when stocks have low earnings or negative earnings. An
alternative statistic defined as the ratio of the difference between stock price and earnings to the
stock price has a valid economic interpretation regardless of whether earnings are negative,
positive or near zero.

The apparent problem with PE ratios and the potential advantage of the use of (P-E)/P are
demonstrated in table one below.

Comparing Two Valuation Measures

Electroniccopy
Electronic copyavailable
available at:
at: [Link]
[Link]
 P EPS PE (P-EPS)/P
10.00 0.50 20.00 0.95
10.00 0.40 25.00 0.96
10.00 0.30 33.33 0.97
10.00 0.20 50.00 0.98
10.00 0.10 100.00 0.99
10.00 0.00 #DIV/0! 1.00
10.00 -0.10 -100.00 1.01
10.00 -0.20 -50.00 1.02
10.00 -0.30 -33.33 1.03
10.00 -0.40 -25.00 1.04
10.00 -0.50 -20.00 1.05

Some observations on differences in the two valuation measures for this example are discussed
below.

First, the impact of a 0.1 change in EPS has a much higher impact on the PE ratio whenever the
EPS figure is near zero. For example, a decrease in earnings from 0.5 to 0.4 causes the PE ratio
to go up by 5 while a decrease from 0.2 to 0.1 results in an increase of 50 points.

Second, the PE ratio is undefined when EPS is 0. When EP is very small the PE ratio has a
large absolute value. An EPS of 0.01 would result in a PE of 500. By contrast, the value of (P-
EPS)/P goes towards 1 when EPS goes to 0.

Third, the PE ratio of a firm with negative earnings near zero is lower than a PE ratio for a firm
with larger negative earnings. This means for negative earnings a decrease in earning (or an
increase in negative earnings) suggests firm valuations have increased, a nonsensical result. By
contrast, the value of (P-E)/P continues to rise even after EPS goes below zero.

Note there is one-to-one relationship between (P-E)/P and P/E. Start with equation one

(𝑝 − 𝑒)
𝑓=
𝑝

Rearrange and solve for p/e to get

𝑝 1
=
𝑒 (1 − 𝑓)

Electronic copy available at: [Link]

if f>1 the PE ratio is negative and undefined.

Section Three: Issues Related to Calculation

of Portfolio PE Ratios

The PE ratio of a portfolio is correctly defined as the sum of the market caps of all firms in the
portfolio divided by the sum of the earnings of all firms in the portfolio.

𝑀𝐶.
𝑃𝐸 =
𝐸.

The PE ratio of a portfolio can be shown to be equal to

1
𝑃𝐸 =
(𝑀𝐶. − 𝐸. )
1− 𝑤.
𝑀𝐶.

where wi is

𝑀𝐶.
𝑤. =
𝑀𝐶.

In an appendix to this paper I show that the two formulas for the PE ratio of a portfolio are
identical.

The PE ratio cannot be calculated by taking the weighted average of firm PE ratios because as
noted in the previous section firm PE ratios are unstable when EPS is near zero and undefined
when below zero.

These methodological issues can best be illustrated with real-world data. Below I have data on
the market capitalization and earnings for four stocks – (1) Apple, (2) Microsoft, (3) Amazon,
and (4) Twitter.

Financial Data for Four Stocks

Market share Market Cap ($ Earnings (P-E)/P
wieight B)

AAPL 0.454 892.16 50.605 0.943

MSFT 0.298 585.37 20.876 0.964
AMZN 0.242 475.37 1.889 0.996
TWTR 0.007 13.11 -0.460 1.035

Electronic copy available at: [Link]

 1.000 1966.0 72.909

The table below confirms that the two methods give the identical value for the PE ratio of the
portfolio.

Results from Two PE Calculation Methods

Sum Market Cap 1966.0

Sum Earnings 72.9

Sum Market Cap/ 26.965

Sum Earnings

Weighted Average (MC-E)/MC 0.963

1/(1-WA) 26.965

The average, weighted by market cap shares, of the four PE ratios in the portfolio provides a
very different estimate for the portfolio PE ratio.

Weighted Average of Firm PE Ratios

market share weight pe

AAPL 0.454 17.63

MSFT 0.298 28.04
AMZN 0.242 251.67
TWTR 0.007 -28.47
1.000

Weighted Average P/E 77.01

Electronic copy available at: [Link]

The weighted average of the firm PE ratios is an incorrect measure of the portfolio PE ratio for
two reasons. The negative PE ratio for Twitter has no economic meaning. Even though Twitter
is a high valuation firm its inclusion in the weighted average serves to reduce the portfolio PE
ratio. Second, Amazon with its high stock price and its small earnings has a disproportionate
impact on the PE ratio of the portfolio. The stock price of Amazon and other firms with
spectacular growth potential is relatively unaffected by earnings. A relatively small decrease
change in current-year Amazon earnings could substantially increase the PE ratio of Amazon
because earnings, the denominator of the fraction, are so small.

In our example, the high PE ratio of Amazon results in an extremely high PE ratio even though
Twitter a firm with a negative PE ratio is included in the calculation of the weighted average.

Note the PE ratio of a portfolio would be undefined if Sum(MCi)<Sum(Ei) or if Sum((MCi-

Ei)/MCi) > 1.0. This could happen for a portfolio of startups where the value of the firm is
determined by expectations of future earnings rather than current earnings.

Section Four: An Analysis of Two

Vanguard Portfolios

The analysis in this section is based on the 40 largest positions in two Vanguard portfolios --
VOT Vanguard mid-cap growth fund, and VOT Vanguard mid-cap value fund. The actual
sample size for VOE is 39 because one firm lacked information on earnings per share. The data
used in this analysis is presented in Appendix Two of this paper.

The chart below contains my calculations of the PE ratio for my samples and the Vanguard
estimate for the PE of the fund.

PE Estimates of Two Vangaurd Funds

VOT VOE
Count 40 39
PE Ratio 45.8 25.0
Method 1

Wa((mc-e)/mc)) 0.9782 0.9601

PE Ratio 45.8 25.0

Method 2

Electronic copy available at: [Link]

 PE Reported by Vanguard 30.6 18.6
for all stocks in the ETF

Ratio of PE for sample to 1.50 1.35

Vanguard fund estimate

Observe once again that the two methods of obtaining a PE ratio for a portfolio provide an
identical result.

Vanguard does not use the methods presented in this paper.

When I clicked on the words PE ratio on the Vanguard chart I obtained the following
explanation.

price/earnings ratio
The price per share of a stock divided by its per-share earnings over the past year. For a
portfolio, the ratio is the weighted average price/earnings ratio of the stocks it holds.

Vanguard did not explain whether it included firms with negative PE ratios in their calculations
or their treatment of firms with extremely large and unstable PE ratios because earnings were
near zero.

My calculation is based on samples of the top positions. Vanguard with easy and complete
access to data in their funds calculated the PE ratio of the entire funds. We are admittedly
comparing apples to oranges.

In theory, how should the omission of smaller positions from my sample bias this comparison?
Smaller firms tend to be startups with low earnings and high expectations so I expect their
omission from my sample will reduce the number of firms with high values of (MC-E)/MC.
This will include both firms with high PE ratios and negative or undefined PE ratios.

My estimate of the PE ratio for the top positions in VOT is 50 percent higher than Vanguard’s
estimate for the PE ratio of the entire fund. My estimate of the PE ratio for the top forty
positions in VOE is 35 percent higher than Vanguard’s estimate for the entire fund.

Electronic copy available at: [Link]

This result does not surprise me. Vanguard states their ETF is weighted value of firm PE ratios.
However, firm PE ratios are unstable when earnings are near zero and nonsensical when
negative.

Section Five: Statistical Inference and Firm

Valuation Measures

Often analysts and researchers create and test hypothesis’ about means or the distribution of
financial ratios. This section employs data from the VOT and VOE funds to demonstrate that the
use of (MC-E) to MC ratio will provide much more meaningful and useful results than the use of
PE ratios.

Below are the descriptive statistics obtained from the top positions in ETFs VOT and VOE.

Descriptive Statistics 40 Positions in VOT and VOE

VOT VOE
PE (MC-E)/MC PE (MC-E)/MC
count 33 40 36 39

Min 18.1 0.945 8.2 0.878

25th 27.3 0.966 16.7 0.942

50th 33.6 0.975 21.0 0.953

75th 48.8 0.986 27.6 0.971

Max 494.9 1.070 106.8 1.105

mean 54.8 0.980 27.0 0.960

std 83.5 0.026 18.9 0.037

The source of data for this chart is a web site named [Link] One firm in VOE
lacked information on earnings so the sample is 40 firms for VOT and 39 firms for VOE.
Several firms had negative earnings and undefined PE ratios.

The (MC-E)/MC statistic has several clear advantages over the traditional PE ratio.

Electronic copy available at: [Link]

First, several firms (17.5 percent of firms in VOT and 7.7 percent of firms in VOE) had negative
earnings and had to be removed from the sample when the PE ratio is employed. Firms with
positive value despite having no current earnings are high-valuation firms. Their entire value
stems from future earnings. The omission of negative PE firms from this analysis artificially
reduces portfolio valuation. These firms have high values of (MC-E)/MC. The use of (MC-
E)/MC allows us to include all firms when comparing portfolios.

Second, the PE ratios are highly positively skewed, as evidenced by thelargest PE ratio – 495 for
VOT and 107 for VOE. The positive skew occurs because of a few firms with small positive
earnings. The positive skew means the PE data is not normally distributed, which limits the
power of some conventional hypothesis tests. The outliers increase the standard deviation,
which decreases the likelihood of rejecting a null hypothesis when a parametric test is conducted
with the PE ratio variable. By contrast, the variable (MC-E)/MC appears symmetric about the
mean and is likely normally distributed.

The VOT and VOE data is used to consider two theories. The first hypothesis is the that the
mean PE ratio of stocks in VOT and VOE are identical. The second hypothesis is that the mean
of (MC-E)/MC are identical. The results are presented below.

Test Results for Differences

Between VOT and VOE

Test p-value
ttest PE ratios 0.943
ttest (mc-e)/mc 0.010

The test for differences in the average PE ratio for the two portfolios is not significant. The test
for differences in the mean value of (MC-E)/MC is highly significant.

A more important test involves the difference between Vanguard valuation measures and the
ones I obtained from the sample of the largest positions from the two ETFs. In order to conduct
this test, I convert the Vanguard PE ratio to its corresponding (MC-E)/MC. (Extend the algebra
shown in Section Two to show that (1-e/p)=(MC-E)/MC.) The results of the test of the null
hypothesis that the mean valuation from the samples of firms are equal to the Vanguard
valuations are presented in the table below.

Electronic copy available at: [Link]

 Testing for Differences Between Vanguard Valuation
and Sample Valuation

VOT VOE
Vanguard PE ratio 30.5 18.6

WA (MC-E)/MC 0.9672 0.9462

consistent with Vanguard
PE

Observed (MC-E)/MC 0.9796 0.9603

STD of (MC-E)/MC 0.0258 0.0373

t 3.0 2.4
P-value 0.0042 0.0235

The valuation of the top positions in VOT and VOE (measured by (MC-E)/MC)) are both higher
than the valuations reported by Vanguard for the entire fund.

Typically, a non-parametric test is employed when data is highly skewed. However, the high
level of skew is not the only problem with the PE ratio. The variable is undefined and
nonsensical when earnings are negative. This problem can’t be fixed by picking a non-
parametric procedure. Fortunately, there is a one-to-one relationship between (MC-E)/MC and
tests can be conducted with this alternative statistic.

Section Five: Concluding Remarks

Every day I turn on CNBC and hear an analyst say something like “Market valuations are a bit
high but the markets PE of 18 or 20 is only a bit above the historic norms and since interest rates
are low and the economy is strong the bull market probably has a couple of more years to run.”
Maybe the market does have a few more years to run. However, many highly diversified funds
with stocks making up a significant share of the market currently have many stocks with market
caps substantially higher than current year earnings. Under current regulations, market analysts
are free to use and report portfolio valuation measures of their choice.

The financial statistics reported on widely traded ETFs and mutual funds are misleading. In my
view regulators need to require the industry to present more accurate and complete financial

Electronic copy available at: [Link]

information on fund holding. This problem could also be fixed by an aggressive and
independent financial press.

Electronic copy available at: [Link]

 Appendix One: Deriving the Relationship
Between the Two PE Formulas

Theorem: The following relation holds.

012 4
= 89 :; =
32 ( 45 62 ∗ 2 2 )
8<2

where wi are mci/sum(mci)

Proof: I expand the term on the right side of the equation for a two-stock portfolio.

Note that wi*(MCi-Ei)/MCi is equal to (MCi-Ei)/(SUM(MCi) for all i.

Note also that 1= SUM(MCi)/SUM(MCi)

Plug both terms into the right side of the equation above and combine like terms.

1
𝑀𝐶. − 𝑀𝐶. − − 𝐸.
𝑀𝐶.

This simplifies to the sum of the market caps to the sum of the earnings in the portfolio, which is
the left side of our equation.

Electronic copy available at: [Link]

 Appendix Two:

Data for VOT

no. ETF Market Mkt Earnings PE Ratio (MC-
Cap. Sjhare E)/MC

1 LRCX 31.56 0.0401 1.5014 21.03 0.952

2 FISV 26.65 0.0339 0.8608 30.97 0.968
3 APH 26.5 0.0337 0.9072 29.22 0.966
4 ROP 25.61 0.0325 0.6827 37.52 0.973
5 ADSK 26.02 0.0331 0.0000 NA 1.000
6 BCR 23.58 0.0300 0.5536 42.6 0.977
7 MCO 27.53 0.0350 0.4794 57.43 0.983
8 EW 23.24 0.0295 0.6969 33.35 0.970
9 CERN 24.04 0.0305 0.6565 36.63 0.973
10 RHT 21.43 0.0272 0.2955 72.53 0.986
11 DLTR 21.68 0.0275 0.9261 23.41 0.957
12 CXO 19.84 0.0252 0.6662 29.79 0.966
13 NOW 20.95 0.0266 -0.1656 NA 1.008
14 MCHP 21.34 0.0271 0.4515 47.27 0.979
15 INCY 24.36 0.0309 -0.1601 NA 1.007
16 SWKS 19.72 0.0251 0.9571 20.61 0.951
17 DLR 25.2 0.0320 0.5154 48.9 0.980
18 WCN 18.5 0.0235 -0.9484 NA 1.051
19 SBAC 17.87 0.0227 0.0361 494.87 0.998
20 EXPE 20.71 0.0263 0.2972 69.7 0.986
21 EQIX 35.98 0.0457 0.2144 167.85 0.994
22 ERTS 5.05 0.0064 -0.3536 NA 1.070
23 ESS 17.25 0.0219 0.5134 33.6 0.970
24 MTD 17 0.0216 0.4164 40.83 0.976
25 BMRN 16.39 0.0208 -0.1787 NA 1.011
26 LVLT 19.39 0.0246 0.6382 30.39 0.967
27 FRC 15.17 0.0193 0.6306 24.06 0.958
28 MGM 17.54 0.0223 0.9722 18.05 0.945
29 MHK 19.23 0.0244 0.9603 20.03 0.950
30 O 15.65 0.0199 0.3207 48.8 0.980
31 CNC 16.18 0.0206 0.7969 20.31 0.951

Electronic copy available at: [Link]

32 AME 15.68 0.0199 0.5264 29.79 0.966
33 VMC 15.48 0.0197 0.3995 38.75 0.974
34 INFO 7.88 0.0100 0.1801 43.77 0.977
35 GPN 15.52 0.0197 0.2606 59.57 0.983
36 FLT 14.94 0.0190 0.4648 32.15 0.969
37 WAT 14.44 0.0183 0.5292 27.29 0.963
38 WDAY 14.42 0.0183 -0.2439 NA 1.017
39 KMX 13.74 0.0175 0.6636 20.71 0.952
40 IDXX 13.96 0.0177 0.2539 54.99 0.982

Electronic copy available at: [Link]

 Data for VOE
ETF Market Cap. Mkt Sjhare Earnings PE Ratio "(mc-E)/mc

1 WDC 25.32 0.0363 0.3955 64.02 0.9844

2 COL 21.9 0.0314 0.8237 26.59 0.9624
3 RCL 27.99 0.0402 1.5665 17.87 0.9440
4 NEM 20.45 0.0293 0.1916 106.76 0.9906
5 WEC 20.67 0.0297 0.9625 21.48 0.9534
6 MTB 24.94 0.0358 1.2976 19.22 0.9480
7 HIG 20.33 0.0292 0.6666 30.5 0.9672
8 KEY 19.78 0.0284 1.0765 18.38 0.9456
9 NWL 20.55 0.0295 1.2446 16.52 0.9394
10 DTE 19.76 0.0284 1.0440 18.93 0.9472
11 WLTW 21.73 0.0312 0.4840 44.9 0.9777
12 FCX 21.61 0.0310 1.2882 16.78 0.9404
13 ES 19.56 0.0281 0.9790 19.99 0.9500
14 PFG 19.53 0.0280 1.2717 15.36 0.9349
15 SYMC 19.67 0.0282 -0.4117 NA 1.0209
16 CFG 18.36 0.0263 1.1713 15.68 0.9362
17 RF 17.66 0.0253 1.1351 15.56 0.9357
18 CLX 17 0.0244 0.6897 24.65 0.9594
19 DVMT 16.45 0.0236 -1.7305 NA 1.1052
20 LNC 16.56 0.0238 1.4511 11.42 0.9124
21 DPS 16.24 0.0233 0.7578 21.44 0.9533
22 ALB 15.4 0.0221 0.2938 52.41 0.9809
23 LH 15.24 0.0219 0.7267 20.98 0.9523
24 AWK 15.34 0.0220 0.4725 32.47 0.9692
25 TAP 16.14 0.0232 1.9692 8.2 0.8780
26 IVZ 15.1 0.0217 0.9346 16.16 0.9381
27 BLL 14.77 0.0212 0.2340 63.13 0.9842
28 CE 14.34 0.0206 0.8197 17.5 0.9428
29 HBAN 15.05 0.0216 0.7738 19.46 0.9486
30 ETR 15.19 0.0218 -0.8924 NA 1.0587
31 AEE 14.79 0.0212 0.6916 21.39 0.9532
32 DHI 15.53 0.0223 0.9994 15.54 0.9356
33 DOV 14.65 0.0210 0.6231 23.52 0.9575
34 BBY 16.35 0.0235 1.1219 14.58 0.9314
35 WRK 15.14 0.0217 0.4242 35.69 0.9720
36 MKL 14.86 0.0213 0.4022 36.95 0.9729

Electronic copy available at: [Link]

37 MSI 14.46 0.0207 0.6117 23.64 0.9577
38 LLL 14.56 0.0209 0.6940 20.99 0.9523
39 CAG 13.97 0.0200 0.5676 24.62 0.9594

Electronic copy available at: [Link]

i
The footnotes in Vanguard ETFS state the firm takes a weighted average of firm PE ratios when
constructing their ETF PE ratios. This post by Dave Nadig discusses some issues with ETFs
reported by the industry. [Link]
[Link]?nopaging=1 This post by Paul Britt discusses some issues with negative PE ratios.
[Link]
[Link]?fullart=1&start=2

Electronic copy available at: [Link]