The Informativeness of Embedded Value Reporting To Stock Price

Accounting & Finance 61 (2021) 5341–5376
The informativeness of embedded value reporting to

stock price
Derrick W. H. Funga , David Joub, Ai Ju Shaoc, Jason J. H. Yehd

a
The Hang Seng University of Hong Kong, Hong Kong
b
Taikang Life Insurance Company Limited, Beijing, China
c
The Ming Chuan University, Taipei, Taiwan
d
The Chinese University of Hong Kong, Hong Kong, China
Abstract
This paper examines the informativeness of embedded value reporting to stock

price by investigating the cross-sectional variations in life insurers’ price to
embedded value ratios. By conducting variance decomposition analysis on a
dataset provided by Morgan Stanley, we find that 15 percent (40 percent) of the
difference between embedded value and stock price can be explained by growth
opportunities and future stock returns in the short (long) run. One-third and
two-thirds of the unexplained variation are attributed to firm- and country-
specific factors, respectively. The above findings provide investors with a better
understanding of the value relevance of embedded value reporting.
Key words: Embedded value accounting; Price to embedded value ratio;

Variance decomposition; Informativeness; Insurer valuation
JEL classification: M41, G22
doi: 10.1111/acfi.12761
1. Introduction
Although price-to-book (P/B) and price-to-earnings (P/E) ratios are com-

monly used in valuing listed firms, they are rarely used in valuing listed life
We would like to thank Grantley Taylor, Imran Haider, Han Fan Ying, Man Lai Tang,
Fei Lung Yuen, Wing Yan Lee, the seminar participants at the 2nd Global Chinese
Accounting Summit, and the workshop participants at the Hang Seng University of
Hong Kong for helpful comments. All errors are our own.
Please address correspondence to Derrick W. H. Fung via email: [email protected]
© 2021 Accounting and Finance Association of Australia and New Zealand

5342 D. W. H. Fung et al./Accounting & Finance 61 (2021) 5341–5376
insurers. The problem with book value and earnings prepared under Interna-
tional Financial Reporting Standards (IFRS) is the delayed recognition and
distribution of profit, resulting in mismatches between accounting figures and
economic performance (Horton, 2007). The problem is amplified for life
insurers that issue insurance policies with several decades of duration. With the
objective of measuring the market-consistent profit and firm value of life
insurers, the European Insurance CFO Forum proposed the accounting
measures for embedded value reporting in the early 2000s. In general,
embedded value can be viewed as the realistic consolidated value of
shareholders’ interests in life insurers and should be closely related to their
stock price, which is the value assigned by stock market participants. After
nearly two decades of development on embedded value reporting standards,
approximately 100 life insurers around the globe have disclosed their embedded
values on a voluntary and unregulated basis (Jacob et al., 2017).
Previous empirical studies have found that embedded value has strong
explanatory power for the stock price and market capitalisation of life insurers
(e.g., Prefontaine et al., 2009; Wu and Hsu, 2011; El-Gazzar et al., 2015;
Zimmermann et al., 2015). Some academic studies, such as Malafronte et al.
(2018), use embedded value as a proxy for firm value. From the industry
perspective, practitioners consider embedded value as an important metric in
evaluating mergers and acquisitions, as it measures life insurers’ performance,
financial strength and firm value (American Academy of Actuaries, 2009). Both
academics and practitioners have the general perception that embedded value is
closely related to stock price. However, we still observe substantial variations in
cross-sectional price to embedded value (P/EV) ratios (i.e., market value of
equity divided by embedded value). For example, at the end of 2017, the P/EV
ratio of Ping An Insurance Company (1.58) was higher than that of China
Taiping Insurance Holdings Company Ltd. (0.84) by 88 percent. A survey
conducted by Milliman (2017) also reveals that the P/EV ratios of European
life insurers range from 0.61 to 1.82. Such substantial variation in P/EV ratios
has not, to the best of our knowledge, been academically studied. There is no
theory asserting whether stock price should equal embedded value, nor
predicting whether P/EV ratios should converge or diverge. Against this
backdrop, how informative are embedded value figures to investors? When a
life insurer’s current P/EV ratio is high, will its future stock return be low or
future growth be high? Or will its P/EV ratio continue to stay high in the
future? These questions motivate us to examine the relationship between
embedded value and stock price in this paper. Understanding the cross-
sectional variations in P/EV ratios is also important to practitioners who base
their mergers and acquisition decisions on voluntary and unregulated embed-
ded value statements.
The difference between embedded value and market capitalisation can be
attributed to factors that are only reflected in market capitalisation, but not in
embedded value statements. According to the classic Gordon Growth Model

D. W. H. Fung et al./Accounting & Finance 61 (2021) 5341–5376 5343
(Gordon, 1959) and subsequent dividend and valuation models (Campbell and
Shiller, 1988; Jiang and Lee, 2007), investors assign higher value to life insurers
with more growth opportunities and larger potential distributed dividends,
resulting in higher P/EV ratios. In such cases, the P/EV ratios for these insurers
will decrease once the extraordinary growth period has passed and distributed
dividends decrease. In addition, a branch of literature in behavioural finance
suggests that the market is inefficient and investors are either too optimistic or
too pessimistic about firms’ future cash flows (e.g., Blasco et al., 2012; Milian,
2015; Birru, 2018; Engelberg et al., 2018). In such cases, the P/EV ratio for life
insurers with more optimistic (pessimistic) future cash flows are higher (lower)
but will gradually decrease (increase), accompanied by lower (higher) stock
returns, once the market realises the actual cash flows.
This study analyses the extent to which the above-mentioned components
contribute to cross-sectional variations in P/EV ratios. Inspired by the variance
decomposition approach used by Vuolteenaho (2002), Callen and Segal (2004),
Callen (2009), Chue (2015), Goh et al. (2017) and Henry (2018) to study the
time-series properties of accounting figures, we mechanically decompose the
cross-sectional variations in life insurers’ P/EV ratios into four components: (1)
future P/EV ratios, (2) future growth of embedded value, (3) future dividend
rates and (4) future stock returns. The effects of above-mentioned growth
opportunities, future dividends and future stock prices are reflected in
components (2), (3) and (4), respectively, with the unexplained variations
reflected in component (1). One of the advantages of the variance decompo-
sition approach is that the four components above fully explain the cross-
sectional variations in P/EV ratios. In other words, the variance decomposition
approach makes sure that no other factors, except for the four components
above, affect the cross-sectional variations in P/EV ratios. As embedded values
are not available in common financial databases, this study is based on a
proprietary panel dataset, provided by Morgan Stanley, which consists of life
insurers from 2002 to 2017. We empirically analyse the relationships between
cross-sectional variations in current P/EV ratios, future P/EV ratios, future
growth of embedded value, future dividend rates and future stock returns.
Based on the variance decomposition approach, we find that in the short run
(i.e., within a 1-year horizon), around 85 percent of the cross-sectional
variation in current P/EV ratios is explained by future P/EV ratios. In other
words, there is strong persistency in P/EV ratios that cannot be explained by
growth opportunities, future dividends and future stock prices. Life insurers
with high (low) P/EV ratios are likely to have high (low) P/EV ratios in the next
year, after controlling for other factors. In addition, around 10 percent of the
cross-sectional variation in current P/EV ratios can be explained by future
stock returns in the short run. Life insurers with high (low) P/EV ratios are
likely to have low (high) stock returns in the next two years. Further analysis
reveals that the negative effect of P/EV ratios on future stock returns is driven
primarily by life insurers from China. In addition, in the long run (i.e., within a

5-year horizon), around 60 percent of the cross-sectional variation in current P/

EV ratios is positively associated with future P/EV ratios, and another 30
percent is positively associated with future growth in embedded value. The
persistency in P/EV ratios gradually decreases as the projection horizon
increases, while the relative importance of future growth in embedded value
gradually increases, suggesting that investors price growth opportunities into
life insurers’ market valuations.
In addition to our baseline empirical findings, we conduct robustness checks.
First, we decompose the persistency in P/EV ratios into two components: one
affected by firm-specific factors and another affected by country-specific
factors. Empirical analysis reveals that firm-and country-specific factors
account for around one-third and two-thirds of the persistency in P/EV ratios,
respectively. Second, as the risk-free interest rate and risk premium may affect
stock returns and the growth of embedded values, we subtract them from
future stock returns and growth of embedded values. With the adjusted future
stock returns and growth of embedded values, we repeat the variance
decomposition analysis and find results similar to the baseline findings. Third,
as more reporting standards have been proposed by professional bodies in the
recent decade, it is possible that the informativeness of embedded value
reporting is time varying. To control for time-varying factors, we repeat the
variance decomposition analysis using cross-sectional demeaned quantities and
find no material difference from the baseline results. Finally, to examine
whether the baseline results are driven by any specific jurisdiction, we repeat the
analysis excluding life insurers from one jurisdiction at a time and find that the
relationship between P/EV ratio and future stock returns is driven primarily by
life insurers from China.
This study contributes to the existing literature on the information
intermediation function of embedded value reporting. We add to the literature
by examining the relationship between embedded value and stock price. While
the existing literature focuses on embedded value’s explanatory power for stock
price (e.g., El-Gazzar et al., 2015; Prefontaine et al., 2009; Wu and Hsu, 2011;
Zimmermann et al., 2015) and how embedded value reporting decreases
information asymmetry (Serafeim, 2011), this study focuses on investigating
why embedded value differs from stock price. Our findings also have important
implications. First, the finding that around 10 percent of the cross-sectional
variation in P/EV ratios is negatively associated with stock returns in the short
run may be indicative of investors’ biased expectations and market inefficiency.
This relationship is driven primarily by Chinese life insurers and builds a
foundation for further asset pricing and behavioural finance studies involving
the P/EV ratio as a potential anomaly candidate for the Chinese stock market.
Second, as the P/EV ratio is often used as a reference for merger and
acquisition decisions as well as stock recommendations for life insurers, this
study provides industry practitioners with an in-depth understanding of cross-

sectional variations in P/EV ratios that can facilitate investment decision-

making.
The remainder of this paper is organised as follows. In Section 2, we survey
the relevant embedded value literature. In Section 3, we discuss the data,
methodology of the variance decomposition approach and baseline empirical
model. We present the empirical results and explore the robustness of our
model in Section 4. In Section 5, we conclude and discuss the implications of
our findings.
2. Literature review
Although Horton et al. (2013) suggest that IFRS improves the information
environment, industry practitioners have expressed concerns over the applica-
bility of IFRS to measure the economics of the life insurance industry. The
problem with IFRS is the delayed recognition of revenues and expenses,
resulting in mismatches between accounting figures and economic performance.
In addition, assets are valued according to fair value, whereas liabilities are
valued according to estimated exit values, resulting in valuation misalignment.
Moreover, statutory reporting also fails to measure the economics of life
insurers as Horton (2007) points out that statutory accounting regulations are
based on conservative accounting rules that undervalue the equity of insurers.
In this connection, the European Insurance CFO Forum launched embedded
value principles in May 2004 (CFO Forum, 2004), June 2008 (CFO Forum,
2008) and October 2009 (CFO Forum, 2009) with the aim of developing a set of
accounting principles recognising life insurers’ profit as their underlying risk
emerges and allowing for a realistic assessment of liabilities. Subsequently,
many life insurers have chosen to supplement their financial statements with
embedded value reporting, which better captures the economics of the
insurance industry.
According to the European Insurance CFO Forum (2009), the embedded
value of a life insurer is a ‘measure of the consolidated value of shareholders’
interests’. It consists of three components: required capital, free surplus and the
value of in-force business. Required capital is the minimum capital that
regulators require to be maintained and it cannot be distributed to sharehold-
ers. Free surplus is the remaining equity whose distribution to shareholders is
not restricted by regulators. The value of in-force business is the risk-adjusted
present value of future shareholder cash flows arising from in-force business.
Year-to-year changes in embedded value are usually due to investment income,
value added by new business, dividends paid and unexpected changes (such as
difference between assumption and actual experience, change in embedded
value assumptions, capital injection, share repurchase, etc.) during the year.
The main elements of embedded value are summarised in Figure 1. In general,
embedded value reporting has more market-consistent elements. Examples

Unexpected
Value Dividends changes Value of
added by paid (year t) in-force
new (year t) business
business (year-end
(year t) of t)
Investment
income
(year t)
Embedded value (year-end of t)

Value of
in-force Free
business surplus
(year-end (year-end
Embedded value (year-end of t-1)
of t-1) of t)
Free
surplus
(year-end
of t-1)
Required
capital
Required (year-end
capital of t)
(year-end
of t-1)
Figure 1 Elements of embedded value. [Colour figure can be viewed at wileyonlinelibrary.com]
include the market valuation of assets, use of market spot rate to discount
future claims and consideration of future policy renewals.
As embedded value reporting measures the value of shareholders’ interest,
embedded values should be closely related to stock prices. In general, previous
studies suggest that embedded value reporting has explanatory power for stock
prices and decreases information asymmetry between life insurers and
investors. One of the earliest studies on embedded value reporting by Serafeim
(2011) suggests that life insurers that adopt embedded value reporting
experience a decrease in information asymmetry. With a sample of Canadian
life insurers, Prefontaine et al. (2009) find significant association between
embedded value and life insurers’ market value of equity, suggesting that
embedded value reporting provides value relevance to external stakeholders.
Wu and Hsu (2011) study a sample of life insurers listed on the Taiwan,
London and Euronext exchanges. They find that embedded value reporting has
an incremental role in stock price valuation and the market value of equity. El-
Gazzar et al. (2015) examine a sample of cross-listed US life insurers and find
that embedded value reporting has incremental explanatory power for stock

price beyond that of US GAAP accounting. Also, Zimmermann et al. (2015)

find that embedded value is superior to financial accounting in estimating the
risk premiums of life insurers.
Despite the tendency to disclose embedded values, many industry practi-
tioners criticise embedded value reporting for opaqueness and giving manage-
ment too much discretion over the underlying assumptions (PwC, 2007).
Vedani et al. (2017) also express concern that different embedded value models
generate different values for life insurers and year-to-year changes in embedded
value often result from model changes. Although the European Insurance CFO
Forum, formed by 20 European insurers, issued guidelines in the last decade
with an objective of standardising embedded value reporting, its members
represent only about 20 percent of the insurers reporting embedded values
worldwide (Jacob et al., 2017). In general, the lack of comparability and
opaqueness of embedded values is a concern shared by many industry
practitioners (e.g., Frasca and LaSorella, 2009; PwC, 2009).
Although previous studies show that embedded values have explanatory
power for stock prices, we still observe substantial variation in P/EV ratios in
the market. How much of the difference between embedded value and stock
price is attributed to growth opportunities, future dividends and future stock
returns? How much of the difference is persistent and unexplained? We will
quantify the extent to which each of the above-mentioned components
contributes to cross-sectional variations in P/EV ratios using the variance
decomposition approach, with its methodology explained in the next section.
3. Data and methodology
As embedded values are not available in commonly used financial databases,

this paper exploits a proprietary panel dataset, provided by Morgan Stanley,
which consists of 21 listed life insurers from seven jurisdictions in the Asia-
Pacific region between 2002 and 2017. Other financial data are downloaded
from Thomson Reuters Datastream. Our study focuses on the life insurers in
the Asia-Pacific region for three reasons. First, it is not a common practice for
life insurers in the United States to disclose their embedded values, making it
difficult to conduct empirical studies on them. Second, the life insurance
premium written in the Asia-Pacific region in 2018 ($US1,092 billion) is higher
than that written in the United States ($US718 billion) and in Europe, the
Middle East and Africa ($US1,009 billion) (Swiss Re, 2019). In other words, we
base our study on insurers from the largest life insurance market in the world.
Third, the accounting conventions in the Asia-Pacific region are likely to be
different from those in Europe, the Middle East and Africa. Basing our study
on insurers from a single region can mitigate the bias caused by heterogeneous
accounting conventions across regions. The life insurers’ names and jurisdic-
tions are reported in Table 1. Embedded value reporting was first proposed in
the early 2000s, and none of the life insurers in our sample disclosed embedded

Table 1
Names of insurers in the sample
Name of insurer Jurisdiction
AIA Group Ltd. Hong Kong

Bangkok Life Assurance PCL Thailand
Cathay Financial Holding Company Taiwan
China Life Insurance Company Ltd. (China) China
China Life Insurance Company Ltd. (Taiwan) Taiwan
China Pacific Insurance (Group) Co., Ltd. China
China Taiping Insurance Holdings Company Ltd. China
Fubon Financial Holding Company Taiwan
Great Eastern Holdings Ltd. Singapore
Hanwha Life Insurance Co., Ltd. South Korea
HDFC Life Insurance Company Ltd India
ICICI Prudential Life Insurance Company Ltd. India
Mercuries Life Insurance Co., Ltd. Taiwan
Mirae Asset Life Insurance Co., Ltd. South Korea
New China Life Insurance Company Ltd. China
Orange Life Insurance Co., Ltd. South Korea
Ping An Insurance Company China
Samsung Life Insurance Co., Ltd. South Korea
SBI Life Insurance Company Ltd. India
Shin Kong Financial Holding Company Taiwan
TONGYANG Life Insurance Co., Ltd. South Korea
This table documents the names of insurers in the panel dataset, provided by Morgan Stanley,
which consists of 21 life insurers from seven jurisdictions in the Asia-Pacific region from 2002
to 2017. We follow Thomson Reuters Eikon’s geographic classification (code: FTSCTRY) to
classify life insurers’ home countries.
values before 2002. It is also interesting to note that embedded value reporting
became more popular after 2008, as half of the life insurers in our sample
started reporting embedded values after that year. The panel data consist of 184
firm-year observations, and the descriptive statistics for embedded value,
market capitalisation, P/EV ratio, annual growth of embedded value, dividend
rate and annual stock return are reported in Panel A of Table 2.
As the number of life insurers that disclose embedded values on a
voluntary basis is limited, the sample size of similar empirical studies on
embedded value is usually small. For example, Prefontaine et al. (2009)
study the relationship between embedded value and market value of equity
based on data from four Canadian life insurers from 2000 to 2008. In the
study by Serafeim (2011) on how embedded value reporting decreases
information asymmetry, 60 firms that disclose embedded values for their life
insurance operations are analysed. Although Wu and Hsu (2011) extend
their sample to include life insurers that disclose embedded values and are
listed on the Taiwan, London and Euronext exchanges, only 122

Table 2
Summary statistics
Percentiles
Variable Mean Std. dev Min 25th 50th 75th Max N
Panel A: Whole sample
Embedded value (in $US millions) 15,673 20,533 459 2,910 7,664 21,342 122,551 184
Market capitalisation (in $US millions) 22,992 34,524 363 2,772 9,215 21,528 194,047 184
P/EV ratio 1.417 1.123 0.214 0.728 1.114 1.688 6.960 184
Annual growth of embedded value 0.188 0.223 –0.292 0.063 0.151 0.242 1.406 163
Dividend rate 0.014 0.013 0 0 0.012 0.022 0.056 184
Annual stock return 0.209 0.524 –0.717 –0.101 0.093 0.337 2.888 163
Mean Mean annual growth Mean annual

P/EV ratio of embedded value Mean dividend rate stock return
Sample period (SD) (SD) (SD) (SD) N
Panel B: Individual insurers
AIA Group Ltd. 2010–2017 1.667 0.107 0.012 0.199 8

(0.246) (0.052) (0.006) (0.187)
Bangkok Life Assurance PCL 2009–2017 1.757 0.212 0.016 0.214 9
(0.526) (0.178) (0.005) (0.397)
Cathay Financial Holding Company 2003–2017 1.225 0.124 0.027 0.121 15
(0.529) (0.133) (0.015) (0.339)
China Life Insurance Company Ltd. (China) 2003–2017 2.171 0.239 0.011 0.343 15
D. W. H. Fung et al./Accounting & Finance 61 (2021) 5341–5376

(1.104) (0.26) (0.006) (0.886)
China Life Insurance Company Ltd. (Taiwan) 2006–2017 0.612 0.264 0.013 0.342 12
(0.141) (0.225) (0.013) (0.564)
(continued)
5349
Table 2 (continued)
5350

China Pacific Insurance (Group) Co., Ltd. 2007–2017 1.791 0.150 0.019 0.214 11
(0.879) (0.125) (0.008) (0.743)
China Taiping Insurance Holdings Company Ltd. 2006–2017 1.788 0.370 0.001 0.394 12
(1.203) (0.247) (0.002) (0.769)
Fubon Financial Holding Company 2006–2017 2.475 0.333 0.031 0.148 12
(1.992) (0.399) (0.017) (0.286)
Great Eastern Holdings Ltd. 2002–2017 0.979 0.091 0.022 0.161 16
(0.159) (0.063) (0.006) (0.273)
Hanwha Life Insurance Co., Ltd. 2010–2017 0.728 0.034 0.007 –0.015 8
(0.135) (0.088) (0.010) (0.167)
HDFC Life Insurance Company Ltd 2017 6.960 N/A 0.003 N/A 1
(N/A) (N/A) (N/A) (N/A)
ICICI Prudential Life Insurance Company Ltd. 2016–2017 3.540 0.161 0.014 0.366 2
(0.377) (N/A) (0.003) (N/A)
Mercuries Life Insurance Co., Ltd. 2012–2017 0.291 0.174 0.005 0.061 6
(0.092) (0.091) (0.011) (0.143)
Mirae Asset Life Insurance Co., Ltd. 2015–2017 0.349 0.065 0.011 0.091 3
(0.008) (0.048) (0.019) (0.133)
New China Life Insurance Company Ltd. 2011–2017 1.400 0.211 0.006 0.231 7
(0.290) (0.069) (0.004) (0.461)
Orange Life Insurance Co., Ltd. 2017 0.860 N/A 0.000 N/A 1
(N/A) (N/A) (N/A) (N/A)
Ping An Insurance Company 2004–2017 1.954 0.300 0.011 0.475 14
(0.966) (0.330) (0.006) (0.813)

Samsung Life Insurance Co., Ltd. 2010–2017 0.772 0.093 0.009 0.044 8
(0.129) (0.116) (0.010) (0.192)
(continued)
Table 2 (continued)

SBI Life Insurance Company Ltd. 2017 4.694 N/A 0.002 N/A 1
(N/A) (N/A) (N/A) (N/A)
Shin Kong Financial Holding Company 2004–2017 0.673 0.137 0.009 0.120 14
(0.358) (0.236) (0.014) (0.484)
TONGYANG Life Insurance Co., Ltd. 2009–2017 0.642 0.084 0.008 –0.001 9
(0.138) (0.135) (0.018) (0.140)
This table presents the summary statistics for the variables used in the variance decomposition analysis. The sample is constructed from a panel
dataset, provided by Morgan Stanley, which consists of 21 life insurers from seven jurisdictions in the Asia-Pacific region from 2002 to 2017.
Panel A reports the summary statistics for the whole sample and Panel B reports the mean of various variables for each life insurer, with the
corresponding standard deviation in brackets. As there is only one firm-year observation for HDFC Life Insurance Company Ltd., Orange Life
Insurance Co., Ltd., and SBI Life Insurance Company Ltd., the data for embedded value growth and annual stock return are not available for
these three life insurers.

5351
observations are available for empirical analysis. El-Gazzar et al. (2015)

analyse 10 listed foreign life insurers on the New York Stock Exchange that
disclose embedded values during the 2000s, but the sample consists of only
76 firm-year observations. A more recent study by Malafronte et al. (2018)
examines whether risk disclosure index for European insurers affects their
embedded values with a sample of 100 firm-year observations. When
compared to other empirical studies on embedded values, our panel data
covers 21 life insurers with 184 firm-year observations, and we consider the
sample size to be appropriate for the purpose of this study.
Substantial cross-sectional variation in P/EV ratios is observed for the whole
sample, with the upper quartile (1.688) greater than the lower quartile (0.728)
by 132 percent. The magnitude of standard deviation (1.123) is similar to that
of the median (1.114), which is consistent with our motivation, stipulated in
Section 1, that substantial cross-sectional variation exists in the market. The
number of firm-year observations for all variables, except annual growth of
embedded value and annual stock return, is 184. The number of observations
for those two variables is smaller than that for the other variables because
2 years’ worth of data are needed to calculate growth of embedded value and
annual stock return.
To gain insight into the statistics of individual life insurers, we report the
mean and standard deviation of the variables for each life insurer in Panel B of
Table 2. The sample periods differ because life insurers started disclosing
embedded values at different times. Three disclosed embedded values only for
2017 and, hence, data for growth of embedded value and annual return are not
available for them. As expected, we observe substantial variation in the P/EV
ratios when we examine the mean for each life insurer. Potential explanations
for such variation are examined in later sections of this paper.
Before analysing the cross-sectional variations in P/EV ratios, we first derive
the cross-sectional link between current P/EV ratio, future P/EV ratio, future
growth of embedded value, future dividend rates and future stock returns. We
begin by expressing the market value of equity (P) to embedded value (EV)
ratio at time t + 1 as follows:
Ptþ1 1 þ ΔPPtþ1 Pt
¼ t
(1)
EVtþ1 1 þ ΔEVEV
tþ1 EV
t
t
where ΔPt+1 = Pt+1 − Pt and ΔEVt+1 = EVt+1 − EVt. Taking natural

logarithms on both sides of Equation (1), we get:

Pt Ptþ1 ΔEVtþ1 ΔPtþ1
ln ¼ ln þ ln 1 þ ln 1 þ (2)
EVt EVtþ1 EVt Pt

To incorporate the dividend components, Equation (2) is re-written as

follows:

Pt Ptþ1 ΔEVtþ1
ln ¼ ln þ ln 1 þ
EVt EVtþ1 EVt
(3)
ΔPtþ1 þ Dtþ1 Dtþ1
ln 1 þ
Pt Pt
where Dt+1 represents the dividend paid during year t + 1. Equation (3) can be
further expressed as follows:

Pt Ptþ1 ΔEVtþ1
ln ¼ ln þ ln 1 þ
EVt EVtþ1 EVt
0 1
Dtþ1

B Pt C ΔPtþ1 þ Dtþ1
B
ln@1 C 1þ
ΔPtþ1 þ Dtþ1 A Pt
1þ
Pt

Ptþ1 ΔEVtþ1
¼ ln þ ln 1 þ (4)
EVtþ1 EVt

Ptþ1 ΔPtþ1 þ Dtþ1
ln ln 1 þ
Ptþ1 þ Dtþ1 Pt

Ptþ1 ΔEVtþ1 Dtþ1
¼ ln þ ln 1 þ þ ln 1 þ
EVtþ1 EVt Ptþ1

ΔPtþ1 þ Dtþ1
ln 1 þ
Pt

Pt
By denoting the natural logarithm of the P/EV ratio ln EV as pevt, growth
t

ΔEVtþ1
of embedded value ln 1 þ EVt as gevt+1, dividend rate ln 1 þ D tþ1
Ptþ1 as dt+1

ΔPtþ1 þDtþ1
and stock return ln 1 þ Pt as rt+1, Equation (4) becomes:
pevt ¼ pevtþ1 þ gevtþ1 þ dtþ1 rtþ1 (5)
Iterating Equation (5) forward by N − 1 periods gives:
N N N
pevt ¼ pevtþN þ ∑ gevtþi þ ∑ dtþi ∑ rtþi : (6)
i¼1 i¼1 i¼1
Based on Equation (6), a high (low) current P/EV ratio can be associated
with a persistent high (low) P/EV ratio in the future, high (low) accumulated

growth of embedded value in the future, high (low) accumulated dividend rates
in the future, and low (high) accumulated stock returns in the future.1
To analyse the cross-sectional variation of P/EV ratios, we decompose the
variance of pevt as follows:
varðpevt Þ ¼ covðpevt ,pevt Þ

N N
¼ covðpevt ,pevtþN Þ þ ∑ cov pevt ,gevtþi þ ∑ covðpevt ,dtþi Þ
i¼1 i¼1 (7)
N
∑ covðpevt , rtþi Þ
i¼1
Dividing both sides of Equation (7) by varðpevt Þ, we have:
N
∑ covðpevt ,gevtþi Þ
covðpevt , pevtþN Þ i¼1
þ
varðpevt Þ varðpevt Þ
(8)
N N
∑ covðpevt ,dtþi Þ ∑ covðpevt ,rtþi Þ
þ i¼1 i¼1 ¼1
varðpevt Þ varðpevt Þ
covðpevt ,pevtþN Þ
Equation (8) can be re-written as follows by denoting varðpevt Þ ,
N N N
∑ covðpevt ,gevtþi Þ ∑ covðpevt ,dtþi Þ ∑ covðpevt ,rtþi Þ
i¼1
, varðpevt Þ , and
varðpevt Þ
i¼1 i¼1
varðpevt Þ as βt ðpev,NÞ, βt ðgev,NÞ, βt ðd,NÞ
and βt ðr,NÞ, respectively:
βt ðpev,NÞ þ βt ðgev,NÞ þ βt ðd, NÞ þ βt ðr,NÞ ¼ 1 (9)
where βt ðpev, NÞ, βt ðgev,NÞ, βt ðd,NÞ and βt ðr,NÞ represent the extent to which
cross-sectional differences in P/EV ratios in year t reflect future P/EV ratios,
future growth of embedded value, future dividend rates and future stock
returns, respectively.
After defining βt ðpev, NÞ, βt ðgev,NÞ, βt ðd,NÞ and βt ðr, NÞ, we can estimate
them by regressing pevj,t+N, ∑N N N
i¼1 gev j,tþi , ∑i¼1 d j,tþi and ∑i¼1 r j,tþi , on pev j,t ,
respectively, for each life insurer j. Alternatively, βt ðpev,NÞ, βt ðgev,NÞ, βt ðd,NÞ
and βt ðr, NÞ can be expressed as the coefficients in the following regressions:
1
In an earlier version of the paper, we developed hypotheses connecting the cross-
sectional variations in current P/EV ratio to future P/EV ratio, future growth of
embedded value, future dividend rates and future stock returns. However, we received
comments and feedback that those hypotheses are tautological statements as the
methodology of variance decomposition clearly establishes the theoretical connection of
the above components. Hence, internal validity exists without formulating hypotheses.

pev j,tþN ¼ βt ðpev,NÞpev j,t þ ɛ j,t ðpev,NÞ

N
∑ gev j,tþi ¼ βt ðgev,NÞpev j,t þ ɛ j,t ðgev,NÞ
i¼1
N (10)
∑ d j,tþi ¼ βt ðd,NÞpev j,t þ ɛ j,t ðd,NÞ
i¼1
N
∑ r j,tþi ¼ βt ðr, NÞpev j,t þ ɛ j,t ðr, NÞ
i¼1
We can perform pooled ordinary least squares (OLS) regressions for the
whole sample period to obtain β^ðpev, NÞ, β^ðgev, NÞ, β^ðd, NÞ and β^ðr,NÞ, which
represent the fraction of cross-sectional variations in P/EV ratios that is
attributable to future P/EV ratios, future growth of embedded value, future
dividend rates and future stock returns, respectively.
4. Empirical results
4.1. Baseline results
We perform regressions on Equation (10) for the whole sample, with the
projected horizon ranging from 1 to 5 years. The estimated coefficients
β^ðpev, NÞ, β^ðgev,NÞ, β^ðd,NÞ and β^ðr, NÞ at the projected horizons are
reported in Table 3. The t-statistics based on the Rogers (1983) standard
error with time clustering and the t-statistics based on the double-clustered
standard error as proposed by Thompson (2011) are also reported for each
estimated coefficient. To visualise how the estimated coefficients evolve over
the projected horizons, we plot the estimated coefficients against the projected
horizons in Figure 2. As more years of data are needed for variance
decomposition analysis with a longer projected horizon, the number of firm-
year observations decreases when the projected horizon increases. In addition,
the sum of the estimated coefficients β^ðpev,NÞ, β^ðgev,NÞ, β^ðd, NÞ and β^ðr,NÞ
equals one, suggesting that the accuracy of log-linear transformation is
evident.
In the short run (i.e., a projected 1-year horizon), most of the cross-sectional
variation in P/EV ratios (85.45 percent) is positively associated with future P/
EV ratios. Life insurers with a high (low) P/EV ratio are likely to also have a
high (low) ratio the next year. Although we observe the proportion of cross-
sectional variation in P/EV ratios attributable to future P/EV ratios gradually
decreases as the projected horizon increases, future P/EV ratios still account for

5356
Table 3
Variance decomposition analysis of the cross-sectional variation in P/EV ratios
Projected horizon Firm-year (1) (2) (3) (4) Sum of

(years) Sample period observations β^ðpev,NÞ β^ðgev,NÞ β^ðd,NÞ β^ðr,NÞ (1)–(4)
1 2002–2016 163 85.45% 4.85% –0.01% 9.71% 1

(10.27)*** (2.98)*** (−0.07) (1.05)
[18.61]*** [1.76]* [−0.07] [1.78]*
2 2002–2015 145 79.48% 11.28% 0.13% 9.11% 1
(14.17)*** (3.96)*** (0.49) (1.56)
[18.19]*** [2.72]*** [0.43] [1.80]*
3 2002–2014 128 79.83% 18.26% 0.46% 1.44% 1
(12.35)*** (4.49)*** (1.51) (0.17)
[15.90]*** [3.21]*** [1.10] [0.21]
4 2002–2013 112 69.73% 21.87% 0.70% 7.70% 1
(7.41)*** (3.63)*** (2.09)* (0.54)
[10.70]*** [3.01]*** [1.23] [0.78]
5 2002–2012 96 59.63% 28.30% 0.61% 11.46% 1
(7.43)*** (4.12)*** (1.46) (0.85)
[10.43]*** [3.34]*** [0.80] [1.21]
This table presents the results of the variance decomposition analysis of the cross-sectional variation in P/EV ratios. The sample is constructed
from a panel dataset, provided by Morgan Stanley, which consists of 21 life insurers from seven jurisdictions in the Asia-Pacific region from 2002
to 2017. The sample period and firm-year observations decrease gradually with the projected horizon, as more years of data are needed for
variance decomposition analysis when the projected horizon increases. β^ðpev, NÞ, β^ðgev, NÞ,β^ðd,NÞ and β^ðr,NÞ represent the proportion of cross-
sectional variation in current P/EV ratios attributable to future P/EV ratios, future growth of embedded value, future dividend rates, and future

stock returns, respectively. The t-statistics based on the Rogers (1983) standard error with time clustering are reported in parentheses, and the t-
statistics based on the double-clustered standard error as proposed by Thompson (2011) are reported in square brackets. *, ** and *** indicate
significance at the 10, 5 and 1 percent levels, respectively.
90.00%
80.00%
70.00%
60.00%
50.00%
40.00%
30.00%
20.00%
10.00%
0.00%
1 2 3 4 5
Projected horizon (in years)
Proporon of variaon aributable to future price to embedded value rao

Proporon of variaon aributable to future growth of embedded value
Proporon of variaon aributable to future stock return
Proporon of variaon aributable to future dividend rate
Figure 2 Variance decomposition analysis of the cross-sectional variation in P/EV ratios. [Colour
figure can be viewed at wileyonlinelibrary.com]
nearly 60 percent of the variation in the long run (i.e., a projected horizon of
5 years). All of the estimated coefficients for β^ðpev,NÞ are statistically and
economically significant, suggesting that P/EV ratios have strong persistency.
As the projected horizon increases from 1 to 5 years, the proportion of cross-
sectional variation in P/EV ratios attributable to future growth of embedded
value gradually increases from 4.85 to 28.30 percent. Life insurers with high
(low) P/EV ratios are likely to experience high (low) growth in the future.
Investors price growth opportunities into life insurers’ valuations, causing
higher P/EV ratios for life insurers with more growth opportunities. All of the
estimated coefficients for β^ðgev, NÞ are statistically and economically significant
regardless of the projected horizon. The above findings are also in line with the
classic Gordon Growth Model (Gordon, 1959), which stipulates that firms with
higher growth rates have higher valuation.
It is also interesting to note that 9.71 and 9.11 percent of the cross-sectional
variation in P/EV ratios is negatively associated with the 1-year and 2-year
stock return, respectively. After double-clustering the standard error as
proposed by Thompson (2011), we find that the estimated coefficients for
β^ðr,NÞ are statistically significant for the 1- to 2-year horizons. Life insurers
with high (low) P/EV ratios are likely to have low (high) stock returns in the

short run. This can be explained by investors’ biased expectations about life
insurers’ future cash flows.2 If investors are too optimistic (pessimistic) about
life insurers’ future cash flows, they assign a higher (lower) valuation to those
life insurers, causing higher (lower) P/EV ratios. When investors realise the
actual cash flows, they correct their expectations, causing lower (higher)
valuations through lower (higher) stock returns. However, the estimated
coefficients for β^ðr, NÞ are not statistically significant for the 3- to 5-year
horizons, making it difficult to draw conclusions in the medium to long run.
Finally, we observe that the estimated coefficients for β^ðd, NÞ are not
statistically or economically significant for most of the projected horizons,
suggesting that the effect of future dividend rates on cross-sectional variations
in P/EV ratios, if any, is trivial.
4.2. Decomposing persistency in P/EV ratios
As noted in Section 4.1, P/EV ratios have strong persistency, which deserves
a closer examination. In this subsection, we decompose the persistency of P/EV
ratios and quantify how much of the persistency is attributable to firm- and
country-specific factors.
We begin by demeaning the pevtþN in Equation (6) by the country average.
Specifically, Equation (6) is modified as follows:
where cpevtþN is the country average of the P/EV ratio at time t + N, and
, which represents the P/EV ratio at time t + N after
demeaning by the country average.
The variance of pevt is decomposed as follows:
2
One may wonder whether there is another possible explanation based on life insurers’
underlying risk. Investors assign lower valuations to riskier life insurers, which may
cause lower P/EV ratios for those life insurers. However, we do not find this argument
convincing because embedded value is market consistent and the calculation of
embedded value incorporates the risks life insurers bear. For example, risk premium
should be added to the discount rate in embedded value reporting so that riskier life
insurers are subject to higher discount rates, resulting in lower embedded values. With
lower market valuation and lower embedded values, the P/EV ratio for riskier life
insurers should not be higher than their peers.

Dividing both sides of Equation (12) by varðpevt Þ, we have:
N N
covðpevt , cpevtþN Þ ∑i¼1 covðpevt , gevtþi Þ ∑i¼1 covðpevt , dtþi Þ
We denote , varðpevt Þ , varðpevt Þ , varðpevt Þ and
N
∑ covðpevt , rtþi Þ
i¼1 varðpevt Þ as , βt ðcpev,NÞ, βt ðgev,NÞ, βt ðd,NÞ and βt ðr,NÞ,
respectively, which can be estimated using the methodology documented in
Section 3. Our coefficients of interest in this subsection are and
βt ðcpev, NÞ, which represent the persistency of P/EV ratios that are attributable
to firm- and country-specific factors, respectively. The empirical results are
documented in Table 4.
The estimated coefficients for β^ðgev,NÞ, β^ðd,NÞ and β^ðr, NÞ in Tables 3 and
4 are identical, as the future growth of embedded value, future dividend rates
and future stock returns are not demeaned by country average. We note that
the summation of and β^ðcpev,NÞ in Table 4 equals β^ðpev,NÞ reported in
Table 3. This is not surprising, as the persistency of P/EV ratios (represented by
β^ðpev, NÞ) can be decomposed into two components, one affected by firm-
specific factors (represented by ) and another affected by country-specific
factors (represented by β^ðcpev,NÞ). As noted in Table 4, around one-third of
the persistency in P/EV ratios is attributable to firm-specific factors and the
remaining two-thirds is attributable to country-specific factors.
Given the empirical results documented in Table 4, a follow-up question
arises: What are the firm- and country-specific factors? The answers to this
question go beyond the scope of this study. However, a discussion on the
potential firm- and country-specific factors could shed light for future studies to
examine the persistency in P/EV ratios.
One of the firm-specific factors could be opaqueness of embedded value
reporting. As no universal accounting standards have been implemented for
embedded value reporting, there could be inconsistent assumptions and
management discretion in embedded value calculations among life insurers
with poor corporate governance (PwC, 2007). Oh et al. (2014) find that loose
internal control over financial reporting regulations leads to poor accounting
information quality. Hammersley et al. (2008) and Kim and Park (2009) report
that material weaknesses lead to negative market reaction. Qi et al. (2017) also
find that poor internal control over financial reporting magnifies managerial
expropriation of corporate resources. Hence, market participants assign lower
market capitalisation to life insurers with more opaque embedded value

5360
Table 4
Variance decomposition analysis when future P/EV ratio is demeaned by country average
(1) (2) (3) (4) (5)

Projected horizon (years) Sample period Firm-year observations β^ðcpev,NÞ β^ðgev,NÞ β^ðd,NÞ β^ðr,NÞ Sum of (1)–(4)
1 2002–2016 163 26.64% 58.81% 4.85% –0.01% 9.71% 1

(9.00)*** (8.84)*** (2.98)*** (−0.07) (1.05)
[6.30]*** [11.07]*** [1.76]* [−0.07] [1.78]*
2 2002–2015 145 25.17% 54.30% 11.28% 0.13% 9.11% 1
(8.77)*** (10.69)*** (3.96)*** (0.49) (1.56)
[6.27]*** [10.32]*** [2.72]*** [0.43] [1.80]*
3 2002–2014 128 24.27% 55.56% 18.26% 0.46% 1.44% 1
(10.24)*** (11.22)*** (4.49)*** (1.51) (0.17)
[5.51]*** [10.03]*** [3.21]*** [1.10] [0.21]
4 2002–2013 112 23.58% 46.15% 21.87% 0.70% 7.70% 1
(11.44)*** (5.78)*** (3.63)*** (2.09)* (0.54)
[4.81]*** [7.01]*** [3.01]*** [1.23] [0.78]
5 2002–2012 96 21.98% 37.65% 28.30% 0.61% 11.46% 1
(12.44)*** (4.84)*** (4.12)*** (1.46) (0.85)
[4.07]*** [5.93]*** [3.34]*** [0.80] [1.21]
This table presents the results of the variance decomposition analysis of the cross-sectional variation in P/EV ratios when the future P/EV ratio is
demeaned by country average. The sample is constructed from a panel dataset, provided by Morgan Stanley, which consists of 21 life insurers
from seven jurisdictions in the Asia-Pacific region from 2002 to 2017. The sample period and firm-year observations decrease gradually with the
projected horizon, as more years of data are needed for variance decomposition analysis when the projected horizon increases. ,
β^ðcpev,NÞ, β^ðgev, NÞ, β^ðd, NÞ and β^ðr, NÞ represent the proportion of cross-sectional variation in current P/EV ratios attributable to future
demeaned P/EV ratios, future country average of P/EV ratios, future growth of embedded value, future dividend rates, and future stock returns,

respectively. The t-statistics based on the Rogers (1983) standard error with time clustering are reported in parentheses, and the t-statistics based
on the double-clustered standard error as proposed by Thompson (2011) are reported in square brackets. *, ** and *** indicate significance at the
10, 5 and 1 percent levels, respectively.
statements, resulting in lower P/EV ratios. This argument is also consistent

with previous studies that suggest poor accounting information quality
(Lambert et al., 2007) and high information risk (Riedl and Serafeim, 2011)
lead to an increase in cost of capital, which has a negative impact on firm value.
Regulatory arbitrage could be one of the country-specific factors. As
regulatory requirements differ across jurisdictions, life insurers are subject to
different regulatory capital requirements, restrictions on investment and
corporate governance standards. As suggested by Durnev and Kim (2005),
Atanasov et al. (2010) and Pham et al. (2020), the effect of regulation is
reflected in market valuation. However, as embedded value does not incorpo-
rate the effect of regulation, P/EV ratios are likely to vary by jurisdiction.
The above explanations are not meant to be exhaustive, and it is not our
intention to assert that they best describe the results in Table 4. More empirical
studies have to be conducted to examine what firm- and country-specific factors
cause the persistency in P/EV ratios.
4.3. Robustness check for future growth of embedded value and future stock
returns
In this subsection, we control for any potential effect that risk-free interest rate
and risk premium have on future growth of embedded value and future stock
returns. When risk-free interest rates are low, the low cost of borrowing facilitates
expansion and growth. This argument is consistent with the literature on interest
rates, monetary policy and growth (e.g., Lin et al., 2018; Grosse-Rueschkamp
et al., 2019). Previous studies also suggest that risk premium and a firm’s growth
rate are associated (e.g., Black and McMillan, 2006; Ashton and Wang, 2013). In
addition, there is a large body of finance literature that suggests risk-free interest
rate and risk premium are associated with stock returns (e.g., Patelis, 1997; Fama
and French, 2002; Maio, 2014; Hau and Lai, 2016). To account for any potential
effect that the risk-free interest rate has on the results of our variance decompo-
sition analysis, we collect risk-free interest rates from Thomson Reuters
Datastream and subtract them from growth of embedded value and stock returns.
The choices of risk-free interest rates for the jurisdictions in our sample are
reported in Table 5. With the adjusted values, we repeat the regressions in
Equation (10) and document the results in Table 6. To control for any potential
effect risk premium has on the results of our variance decomposition analysis, we
repeat the above procedures but replace the risk-free interest rate with risk
premium, which is estimated following Damodaran (1999).3 The empirical
findings are documented in Table 7.
The estimated coefficients for β^ðpev,NÞ and β^ðd, NÞ in Tables 3, 6 and 7 are
identical, as future P/EV ratios and future dividend rates are not adjusted for
risk-free interest rate and risk premium. We observe that there are no material
3
Historical risk premium data can be downloaded from the database maintained by
Damodaran, which is available at http://pages.stern.nyu.edu/~adamodar/.

Table 5
Choice of risk-free interest rates
Jurisdiction Risk-free interest rate (Thomson Reuters Datastream data items in brackets)
China China government bond yield – 1 year (TRCH1YT)

Hong Kong Hong Kong government bond yield – 1 year (TRHK1YT)
India India government bond yield – 1 year (TRIN1YT)
Singapore Singapore government bond yield – 1 year (TRSG1YT)
South Korea Korea government bond yield – 1 year (TRKR1YT)
Taiwan Taiwan government bond yield – 2 years (TRTW2YT)
Thailand Thailand government bond yield – 1 year (TRTH1YC)
This table documents the choice of risk-free interest rates collected from Thomson Reuters
Datastream, which consists of risk-free interest rates from seven jurisdictions in the Asia-
Pacific region from 2002 to 2017. The Datastream codes are reported in brackets. The choice
of risk-free interest rate for Taiwan is Taiwan government bond yield – 2 years because no
shorter duration of government bond is available.
differences between the estimated coefficients for β^ðgev, NÞ and β^ðr,NÞ in

Tables 3, 6 and 7. After adjusting for risk-free interest rate and risk premium,
there is still an increasing trend for the proportion of cross-sectional variation
in P/EV ratios attributable to future growth of embedded value as the projected
horizon increases. In addition, around 10 percent of the cross-sectional
variation in P/EV ratios is negatively associated with the 1-year and 2-year
stock returns. This relationship is not statistically significant for a projected
horizon longer than two years. The above findings suggest that the baseline
empirical results are not biased by risk-free interest rates or risk premiums.
4.4. Robustness check for time-varying factors
In this subsection, we control for the potential effect of time-varying factors

on our baseline results. One may suggest that it takes time for professional
bodies to publish accounting standards for embedded value reporting and that
embedded value figures may be more informative in recent years, leading to less
of the cross-sectional variation in P/EV ratios being attributable to future P/EV
ratios. It is also possible that there is a learning curve for investors to price in
future growth of embedded value, causing more of the cross-sectional variation
in P/EV ratios being attributable to future growth in embedded value in recent
years. To address the potential effect of time-varying factors, we follow Chue
(2015) to demean all variables in Equation (6) cross-sectionally so that the
pooled OLS estimates in the regressions on Equation (10) are equivalent to
estimates from a time-fixed effect model. The variance decomposition analysis
is then repeated with the demeaned variables and the results are reported in
Table 8.

Table 6
Variance decomposition analysis after controlling for risk-free interest rates
(1) (2) (3) (4)

Projected horizon (years) Sample period Firm-year observations β^ðpev,NÞ β^ðgev,NÞ β^ðd,NÞ β^ðr,NÞ Sum of (1)–(4)
1 2002–2016 163 85.45% 4.26% –0.01% 10.30% 1

(10.27)*** (2.72)*** (−0.07) (1.13)
[18.61]*** [1.55] [−0.07] [1.91]*
2 2002–2015 145 79.48% 10.21% 0.13% 10.18% 1
(14.17)*** (3.37)*** (0.49) (1.76)
[18.19]*** [2.40]** [0.43] [2.01]*
3 2002–2014 128 79.83% 16.46% 0.46% 3.24% 1
(12.35)*** (3.93)*** (1.51) (0.39)
[15.90]*** [2.79]*** [1.10] [0.64]
4 2002–2013 112 69.73% 19.43% 0.70% 10.14% 1
(7.41)*** (3.26)*** (2.09)* (0.73)
[10.70]*** [2.57]** [1.23] [1.03]
5 2002–2012 96 59.63% 25.04% 0.61% 14.72% 1
(7.43)*** (3.68)*** (1.46) (1.13)
[10.43]*** [2.83]*** [0.80] [1.52]
This table presents the results of the variance decomposition analysis of the cross-sectional variation in P/EV ratios after controlling for risk-free
interest rates. The sample is constructed from a panel dataset, provided by Morgan Stanley, which consists of 21 life insurers from seven
jurisdictions in the Asia-Pacific region from 2002 to 2017. The sample period and firm-year observations decrease gradually with the projected
horizon, as more years of data are needed for variance decomposition analysis when the projected horizon increases. β^ðpev,NÞ, β^ðgev, NÞ, β^ðd,NÞ
and β^ðr, NÞ represent the proportion of cross-sectional variation in current P/EV ratios attributable to future P/EV ratios, future growth of
embedded value in excess of risk-free interest rates, future dividend rates, and future stock returns in excess of risk-free interest rates, respectively.

The t-statistics based on the Rogers (1983) standard error with time clustering are reported in parentheses, and the t-statistics based on the
double-clustered standard error as proposed by Thompson (2011) are reported in square brackets. *, ** and *** indicate significance at the 10, 5
and 1 percent levels, respectively.
5363
5364
Table 7
Variance decomposition analysis after controlling for risk premium
(1) (2) (3) (4)

1 2002–2016 163 85.45% 4.69% –0.01% 9.87% 1

(10.27)*** (2.75)*** (−0.07) (1.06)
[18.61]*** [1.70]* [−0.07] [1.79]*
2 2002–2015 145 79.48% 11.07%*** 0.13% 9.33% 1
(14.17)*** (3.82)*** (0.49) (1.59)
[18.19]*** [2.66]*** [0.43] [1.83]*
3 2002–2014 128 79.83% 18.10% 0.46% 1.60% 1
(12.35)*** (4.40)*** (1.51) (0.19)
[15.90]*** [3.16]*** [1.10] [0.23]
4 2002–2013 112 69.73% 21.49% 0.70% 8.09% 1
(7.41)*** (3.54)*** (2.09)* (0.57)
[10.70]*** [2.93]*** [1.23] [0.82]
5 2002–2012 96 59.63% 27.47% 0.61% 12.29% 1
(7.43)*** (3.98)*** (1.46) (0.91)
[10.43]*** [3.23]*** [0.80] [1.29]
This table presents the results of the variance decomposition analysis of the cross-sectional variation in P/EV ratios after controlling for risk
premium. The sample is constructed from a panel dataset, provided by Morgan Stanley, which consists of 21 life insurers from seven jurisdictions
in the Asia-Pacific region from 2002 to 2017. The sample period and firm-year observations decrease gradually with the projected horizon, as
more years of data are needed for variance decomposition analysis when the projected horizon increases. β^ðpev,NÞ, β^ðgev,NÞ, β^ðd, NÞ and
β^ðr,NÞ represent the proportion of cross-sectional variation in current P/EV ratios attributable to future P/EV ratios, future growth of
embedded value in excess of risk premium, future dividend rates, and future stock returns in excess of risk premium, respectively. The t-statistics

based on the Rogers (1983) standard error with time clustering are reported in parentheses, and the t-statistics based on the double-clustered
standard error as proposed by Thompson (2011) are reported in square brackets. *, ** and *** indicate significance at the 10, 5 and 1 percent
levels, respectively.
We observe that the estimated coefficients for β^ðpev,NÞ in Table 8 are slightly
higher than those in Table 3. For a projected horizon of 5 years, β^ðpev,NÞ
increases from 59.63 to 68.07 percent after controlling for time-varying factors.
The empirical evidence supporting the persistency in P/EV ratios is even
stronger after controlling for time-varying factors.
While we observe no material differences between the estimated coefficients
for β^ðgev,NÞ in Table 3 and 8, the estimated coefficients for β^ðd,NÞ become
statistically significant in Table 8. However, the largest proportion of cross-
sectional variation in demeaned P/EV ratios attributable to future demeaned
dividend rates is less than 1.5 percent, which is not economically significant.
Hence, we still do not find strong evidence supporting that future dividend rates
have a substantial effect on the cross-sectional variations in P/EV ratios.
It is interesting to note that β^ðr,NÞ becomes statistically insignificant for all
horizons. However, caution should be exercised before drawing any conclusion
regarding the association between P/EV ratios and future stock returns.
Different jurisdictions may experience different stock market cycles in any
particular year and, hence, there is no specific valid reason for demeaning stock
returns cross-sectionally. In this connection, when examining the relationship
between P/EV ratios and future stock returns, we consider the results from
variance decomposition analysis after controlling for risk-free interest rate and
risk premium, as documented in Section 4.3, to be more robust than the results
documented in this subsection.
4.5. Robustness check to examine whether empirical results are driven by any
specific jurisdiction
As noted from Panel B of Table 2, the stock returns of life insurers from
China generally have higher volatility. In addition, most life insurers from
Taiwan and South Korea have P/EV ratios lower than those from other
jurisdictions. Hence, one may be sceptical about whether the baseline results
are driven primarily by jurisdiction. To address this issue, we exclude life
insurers from one jurisdiction at a time and repeat the variance decomposition
analysis to examine if the results are materially different from the baseline
results. The results of this robustness check are reported in Table 9.
Life insurers from China, Hong Kong, India, Singapore, South Korea,
Taiwan and Thailand are excluded from Panels A, B, C, D, E, F and G of
Table 9, respectively. Most of the estimated coefficients for β^ðpev,NÞ and
β^ðgev, NÞ are around 60 and 30 percent, respectively, over a 5-year horizon. The
proportion of variation in current P/EV ratios attributable to future P/EV
ratios gradually decreases as the projection horizon increases. However, the
proportion of variation in P/EV ratios attributable to the future growth of
embedded value gradually increases as the projection horizon increases. All of
the above observations are consistent with our baseline empirical results.

5366
Table 8
Variance decomposition analysis with variables demeaned cross-sectionally
(1) (2) (3) (4)

1 2002–2016 163 90.41% 5.91% 0.27% 3.41% 1

(24.59)*** (2.40)** (2.09)* (0.92)
[32.03]*** [2.39]** [1.56] [1.01]
2 2002–2015 145 82.82% 12.08% 0.59% 4.51% 1
(17.93)*** (2.95)*** (2.83)** (0.81)
[21.28]*** [2.97]*** [1.94]* [0.95]
3 2002–2014 128 79.33% 17.26% 1.03% 2.39% 1
(14.48)*** (3.79)*** (3.71)*** (0.32)
[17.06]*** [3.03]*** [2.31]** [0.35]
4 2002–2013 112 74.28% 23.35% 1.46% 0.91% 1
(11.50)*** (3.76)*** (4.34)*** (0.08)
[12.92]*** [3.06]*** [2.33]** [0.09]
5 2002–2012 96 68.07% 30.64% 1.46% –0.16% 1
(11.61)*** (4.01)*** (3.80)*** (−0.01)
[10.70]*** [3.22]*** [1.76]* [−0.01]
This table presents the results of the variance decomposition analysis of the cross-sectional variation in P/EV ratios with variables demeaned
cross-sectionally. The sample is constructed from a panel dataset, provided by Morgan Stanley, which consists of 21 life insurers from seven
jurisdictions in the Asia-Pacific region from 2002 to 2017. The sample period and firm-year observations decrease gradually with the projected
horizon, as more years of data are needed for variance decomposition analysis when the projected horizon increases. β^ðpev,NÞ, β^ðgev, NÞ, β^ðd,NÞ
and β^ðr, NÞ represent the proportion of cross-sectional variation in current demeaned P/EV ratios attributable to future demeaned P/EV ratios,
demeaned future growth of embedded value, future demeaned dividend rates, and future demeaned stock returns, respectively. The t-statistics

based on the Rogers (1983) standard error with time clustering are reported in parentheses, and the t-statistics based on the double-clustered
standard error as proposed by Thompson (2011) are reported in square brackets. *, ** and *** indicate significance at the 10, 5 and 1 percent
levels, respectively.
Table 9
Variance decomposition analysis after removing insurers from specific jurisdictions
Firm-year (1) (2) (3) (4)

Projected horizon (years) Sample period observations β^ðpev,NÞ β^ðgev,NÞ β^ðd,NÞ β^ðr,NÞ Sum of (1)–(4)
Panel A: Excluding insurers from China
1 2002–2016 109 90.90% 4.90% 0.31% 3.90% 1

(18.09)*** (1.75) (1.26) (1.08)
[18.91]*** [1.35] [1.25] [1.11]
2 2002–2015 96 82.58% 12.31% 0.93% 4.18% 1
(10.91)*** (2.39)** (2.33)** (0.58)
[13.25]*** [1.94]* [2.07]* [0.71]
3 2002–2014 84 78.59% 16.54% 1.93% 2.94% 1
(10.41)*** (3.08)*** (3.80)*** (0.43)
[12.42]*** [2.27]** [3.51]*** [0.45]
4 2002–2013 73 73.29% 24.26% 2.85% –0.40% 1
(7.84)*** (3.43)*** (3.69)*** (−0.04)
[9.34]*** [2.39]** [3.76]*** [−0.04]
5 2002–2012 62 61.50% 28.39% 3.47% 6.65% 1
(5.79)*** (3.52)*** (3.17)*** (0.51)
[8.08]*** [2.38]** [3.52]*** [0.57]
Panel B: Excluding insurers from Hong Kong
1 2002–2016 156 84.88% 5.15% –0.01% 9.99% 1

(10.08)*** (3.03)*** (−0.07) (1.05)
[18.25]*** [1.83]* [−0.07] [1.79]*

2 2002–2015 139 78.64% 11.81% 0.13% 9.42% 1
(14.84)*** (4.21)*** (0.47) (1.61)
[18.19]*** [2.83]*** [0.42] [1.85]*
(continued)
5367
Table 9 (continued)
5368
Firm-year (1) (2) (3) (4)

3 2002–2014 123 79.08% 18.82% 0.46% 1.64% 1

(13.16)*** (4.61)*** (1.48) (0.19)
[16.04]*** [3.31]*** [1.09] [0.24]
4 2002–2013 108 69.10% 22.24% 0.70% 7.96% 1
(7.82)*** (3.63)*** (2.06)* (0.57)
[11.11]*** [3.06]*** [1.22] [0.81]
5 2002–2012 93 59.51% 28.35% 0.61% 11.53% 1
(7.87)*** (4.09)*** (1.46) (0.87)
[10.69]*** [3.36]*** [0.80] [1.22]
Panel C: Excluding insurers from India
1 2002–2016 162 84.88% 4.94% –0.01% 10.19% 1

(10.03)*** (3.01)*** (−0.04) (1.08)
[18.34]*** [1.76]* [−0.04] [1.84]*
2 2002–2015 145 79.48% 11.28% 0.13% 9.11% 1
(14.17)*** (3.96)*** (0.49) (1.56)
[18.19]*** [2.72]*** [0.43] [1.80]*
3 2002–2014 128 79.83% 18.26% 0.46% 1.44% 1
(12.35)*** (4.49)*** (1.51) (0.17)
[15.90]*** [3.21]*** [1.10] [0.21]
4 2002–2013 112 69.73% 21.87% 0.70% 7.70% 1
(7.41)*** (3.63)*** (2.09)* (0.54)
[10.96]*** [3.01]*** [1.23] [0.78]
5 2002–2012 96 59.63% 28.30% 0.61% 11.46% 1

(7.43)*** (4.12)*** (1.46) (0.85)
[10.43]*** [3.34]*** [0.80] [1.21]
(continued)
Table 9 (continued)
Firm-year (1) (2) (3) (4)

Panel D: Excluding insurers from Singapore
1 2002–2016 148 86.11% 4.70% 0.02% 9.17% 1

(10.36)*** (2.89)*** (0.12) (0.99)
[18.73]*** [1.70]* [0.13] [1.67]*
2 2002–2015 131 80.73% 10.80% 0.21% 8.26% 1
(14.55)*** (3.70)*** (0.76) (1.41)
[18.57]*** [2.59]** [0.70] [1.62]
3 2002–2014 115 81.38% 17.21% 0.65% 0.77% 1
(12.65)*** (4.11)*** (2.02)* (0.09)
[16.22]*** [2.99]*** [1.54] [0.11]
4 2002–2013 100 72.05% 20.08% 1.04% 6.84% 1
(7.62)*** (3.14)*** (2.93)*** (0.47)
[11.30]*** [2.69]*** [1.83]* [0.68]
5 2002–2012 85 62.52% 25.48% 1.15% 10.86% 1
(7.80)*** (3.35)*** (2.23)* (0.77)
[10.79]*** [2.90]*** [1.49] [1.08]
Panel E: Excluding insurers from South Korea
1 2002–2016 139 83.63% 3.36% –0.11% 13.13% 1

(9.61)*** (1.85)* (−0.93) (1.40)
[16.52]*** [1.13] [−0.63] [2.19]**
2 2002–2015 125 77.73% 7.72% –0.15% 14.70% 1
(13.28)*** (2.20)** (−0.72) (2.60)**

[15.98]*** [1.71]* [−0.43] [2.58]**
3 2002–2014 112 78.63% 12.75% 0.00% 8.61% 1
(continued)
5369
Table 9 (continued)
5370
Firm-year (1) (2) (3) (4)

(11.23)*** (2.94)*** (0.00) (1.05)

[14.41]*** [2.12]** [0.00] [1.17]
4 2002–2013 99 67.93% 14.22% 0.05% 17.80% 1
(6.99)*** (2.75)*** (0.21) (1.37)
[9.66]*** [1.85]* [0.08] [1.68]*
5 2002–2012 86 56.77% 20.06% –0.15% 23.33% 1
(7.42)*** (3.87)*** (−0.38) (2.18)**
[8.82]*** [2.25]** [−0.17] [2.32]**
Panel F: Excluding insurers from Taiwan
1 2002–2016 109 79.79% 5.68% –0.10% 14.63% 1

(6.70)*** (1.03) (−0.36) (0.88)
[11.46]*** [1.56] [−0.52] [1.56]
2 2002–2015 96 72.29% 13.97% –0.06% 13.80% 1
(11.91)*** (2.83)*** (−0.16) (1.71)
[13.87]*** [2.91]*** [−0.20] [1.89]*
3 2002–2014 84 71.56% 24.22% –0.05% 4.28% 1
(8.04)*** (2.92)*** (−0.10) (0.29)
[10.59]*** [3.33]*** [−0.11] [0.37]
4 2002–2013 73 56.97% 29.77% –0.25% 13.52% 1
(4.63)*** (2.30)** (−0.37) (0.56)
[6.63]*** [3.10]*** [−0.36] [0.86]
5 2002–2012 62 48.54% 41.91% –0.97% 10.52% 1
(5.00)*** (3.01)*** (−1.16) (0.47)

[6.29]*** [3.77]*** [−1.08] [0.75]
(continued)
Table 9 (continued)
Firm-year (1) (2) (3) (4)

Panel G: Excluding insurers from Thailand
1 2002–2016 155 85.02% 4.87% –0.01% 10.13% 1

(9.67)*** (2.80)*** (−0.08) (1.03)
[17.89]*** [1.70]* [−0.09] [1.78]*
2 2002–2015 138 79.08% 11.42% 0.13% 9.38% 1
(13.07)*** (3.86)*** (0.48) (1.48)
[17.83]*** [2.66]*** [0.42] [1.81]*
3 2002–2014 122 79.82% 18.56% 0.47% 1.15% 1
(11.56)*** (4.43)*** (1.47) (0.13)
[15.55]*** [3.16]*** [1.10] [0.16]
4 2002–2013 107 70.17% 22.23% 0.71% 6.90% 1
(7.03)*** (3.59)*** (2.02)* (0.46)
[10.80]*** [2.96]*** [1.22] [0.68]
5 2002–2012 92 59.84% 28.83% 0.61% 10.72% 1
(7.23)*** (4.15)*** (1.46) (0.79)
[10.44]*** [3.34]*** [0.79] [1.12]
This table presents the results of the variance decomposition analysis of the cross-sectional variation in P/EV ratios. The sample is constructed
from a panel dataset, provided by Morgan Stanley, which consists of 21 life insurers from seven jurisdictions in the Asia-Pacific region from 2002
to 2017. Life insurers from China, Hong Kong, India, Singapore, South Korea, Taiwan, and Thailand are excluded from Panels A, B, C, D, E, F
and G, respectively. The sample period and firm-year observations decrease gradually with the projected horizon, as more years of data are
needed for variance decomposition analysis when the projected horizon increases. β^ðpev,NÞ, β^ðgev, NÞ, β^ðd, NÞ and β^ðr, NÞ represent the
proportion of cross-sectional variation in current P/EV ratios attributable to future P/EV ratios, future growth of embedded value, future

dividend rates, and future stock returns, respectively. The t-statistics based on the Rogers (1983) standard error with time clustering are reported
in parentheses, and the t-statistics based on the double-clustered standard error as proposed by Thompson (2011) are reported in square brackets.
*, ** and *** indicate significance at the 10, 5 and 1 percent levels, respectively.
5371
Although the estimated coefficients for β^ðd, NÞ in Panel A are statistically

significant, they are not economically significant. The proportion of the current
P/EV ratio attributable to future dividends is less than 3.5 percent when life
insurers from China are excluded in Panel A. Most of the estimated coefficients
for β^ðd,NÞ in the other panels are economically and statistically insignificant.
Hence, consistent with the baseline results, we do not find convincing evidence
to support that the cross-sectional variation in P/EV ratios is substantially
attributable to future dividends.
The relationship between P/EV ratio and future stock returns is more subtle.
When analysis is conducted on the whole sample in Table 3, we find that
around 10 percent of the variation in P/EV ratios is negatively associated with
stock return in the short run. However, such a relationship vanishes when life
insurers from China are excluded from the analysis in Panel A of Table 9,
implying that life insurers from China are the main drivers of the relationship
between P/EV ratio and stock returns observed in the baseline regression.4 The
empirical evidence in this paper is consistent with the literature (e.g. Chu et al.,
2016) in the sense that investor sentiment has a tremendous impact on stock
returns in the Chinese stock market. Liu et al. (2016) specifically point out that
the Chinese stock market crash of 2015 is due to investors’ overreaction. In
addition, when life insurers from South Korea are excluded from analysis in
Panel E of Table 9, the estimated coefficients for β^ðr,NÞ are statistically
significant at the 1-, 2-, 4- and 5-year horizons. These results suggest that high
(low) P/EV ratios are associated with low (high) stock returns in both the short
and long run, except for life insurers from South Korea. The empirical results
as a whole imply that the relationship between P/EV ratio and future stock
returns are country-specific.
5. Conclusion
As embedded value reporting has become more popular in the life insurance
industry over the last decade, it has gained attention from both investors and
industry practitioners regarding mergers and acquisition decisions and stock-
picking recommendations. Given the substantial cross-sectional variation in P/
EV ratios observed in the market, this study examines the informativeness of
embedded value reporting to stock price and investigates why embedded value
differs from stock price. In particular, we analyse the cross-sectional variation
in P/EV ratios by mechanically decomposing the P/EV ratio into four
4
To examine if any specific life insurers in China drive the negative association between
P/EV ratio and stock returns, we further exclude Chinese life insurers one by one and
repeat the variance decomposition analysis. If China Life Insurance Company Ltd.
(China), China Pacific Insurance (Group) Co., Ltd., or China Taiping Insurance
Holdings Company Ltd. is excluded from the analysis, the above negative association
vanishes. Hence, out of the five Chinese life insurers in our sample, three are the main
drivers of the negative association between P/EV ratio and stock returns.

components: future P/EV ratio, future growth of embedded value, future stock
returns and future dividend rates. Based on our variance decomposition
analysis of Morgan Stanley’s proprietary panel dataset, we find strong
persistency in P/EV ratios, with one-third and two-thirds of the persistency
explained by firm- and country-specific factors, respectively. Life insurers’
market capitalisation tends to keep an enduring ratio with the embedded value.
In addition, we observe that the proportion of variation in P/EV ratios
attributable to the future growth of embedded value gradually increases when
the projected horizon increases. Given its high persistency, the P/EV ratio still
reflects future growth to a certain extent. We also find that in the short run,
variation in P/EV ratios reflects future stock returns. Life insurers with high
(low) P/EV ratios are likely to experience low (high) stock returns in the next
2 years. This relationship is driven mainly by life insurers from China. Further
analyses are also conducted to ensure the above results are robust to risk-free
interest rate, risk premium and time-varying factors.
Our findings have important implications. The high persistency in P/EV
ratios reflects that embedded value does not incorporate all the information
embedded in stock price. Hence, investors should exercise caution when relying
on embedded value as a criterion for life insurer valuation. At least, the
underlying embedded value assumptions should be carefully scrutinised across
life insurers, as they do not include growth opportunities. Future studies can be
conducted to examine the firm- and country-specific factors that cause the high
persistency in P/EV ratios. In addition, as around 10 percent of the variation in
P/EV ratios is negatively associated with stock return in the short run, more
asset pricing studies should be conducted to examine whether the P/EV ratio is
a potential anomaly candidate and whether alpha portfolios can be constructed
based on P/EV ratios, especially for the Chinese life insurers. Finally, given the
relatively few academic studies on cross-sectional variations in P/EV ratios, our
study also lays the groundwork for further studies on embedded value
reporting, a relatively new type of reporting that has gained increasing
attention in the last decade.
References
American Academy of Actuaries, 2009, Embedded value (EV) reporting. Available at:
https://www.actuary.org/sites/default/files/files/publications/Practice_Note_Practice_
note_to_assist_actuaries_working_for_life_insurance_companies_with_the_calcula
tion_of_embedded_values_may2009.pdf
Ashton, D., and P. Wang, 2013, Terminal valuations, growth rates and the implied cost
of capital, Review of Accounting Studies 18, 261–290.
Atanasov, V., B. Black, C. Ciccotello, and S. Gyoshev, 2010, How does law affect
finance? An examination of equity tunneling in Bulgaria, Journal of Financial
Economics 96, 155–173.
Birru, J., 2018, Day of the week and the cross-section of returns, Journal of Financial
Economics 130, 182–214.

Black, A. J., and D. G. McMillan, 2006, Asymmetric risk premium in value and growth
stocks, International Review of Financial Analysis 15, 237–246.
Blasco, N., P. Corredor, and S. Ferreruela, 2012, Market sentiment: a key factor of
investors’ imitative behaviour, Accounting and Finance 52, 663–689.
Callen, J. L., 2009, Shocks to shocks: a theoretical foundation for the information
content of earnings, Contemporary Accounting Research 26, 135–166.
Callen, J. L., and D. Segal, 2004, Do accruals drive firm-level stock returns? A variance
decomposition analysis, Journal of Accounting Research 42, 527–560.
Campbell, J. Y., and R. J. Shiller, 1988, The dividend-price ratio and expectations of
future dividends and discount factors, The Review of Financial Studies 1, 195–228.
CFO Forum, 2004, European insurers move to improve transparency for investors.
Available at: http://www.cfoforum.eu/letters/0405_Press_Release-EEV-VPO.pdf
CFO Forum, 2008, CFO Forum: market consistent embedded value principles. Available
at: http://cfoforum.eu/downloads/CFO-Forum-MCEV-Principles-and-Guidance.pdf
CFO Forum, 2009, CFO Forum: market consistent embedded value principles. Available
at: http://www.cfoforum.eu/downloads/MCEV_Principles_and_Guidance_October_
2009.pdf
Chu, X., C. Wu, and J. Qiu, 2016, A nonlinear Granger causality test between stock
returns and investor sentiment for Chinese stock market: a wavelet-based approach,
Applied Economics 48, 1915–1924.
Chue, T. K., 2015, Understanding cross-country differences in valuation ratios: a
variance decomposition approach, Contemporary Accounting Research 32, 1617–1640.
Damodaran, A. 1999, Estimating equity risk premiums, Finance working paper FIN-99-
021 (New York University). Available at: https://archive.nyu.edu/handle/2451/26918
Durnev, A., and E. H. Kim, 2005, To steal or not to steal: firm attributes, legal
environment, and valuation, The Journal of Finance 60, 1461–1493.
El-Gazzar, S. M., R. A. Jacob, and S. P. McGregor, 2015, The relative and incremental
valuation effects of embedded value disclosure by life insurers: evidence from cross-
listed firms in the US, Accounting Horizons 29, 327–339.
Engelberg, J., R. D. McLean, and J. Pontiff, 2018, Anomalies and news, The Journal of
Finance 73, 1971–2001.
Fama, E. F., and K. R. French, 2002, The equity premium, The Journal of Finance 57,
637–659.
Frasca, R., and K. LaSorella, 2009, Embedded value: practice and theory, Actuarial
Practice Forum, March.
Goh, B. W., C. Y. Lim, G. J. Lobo, and Y. H. Tong, 2017, Conditional conservatism
and debt versus equity financing, Contemporary Accounting Research 34, 216–251.
Gordon, M. J., 1959, Dividends, earnings, and stock prices, The Review of Economics
and Statistics 41, 99–105.
Grosse-Rueschkamp, B., S. Steffen, and D. Streitz, 2019, A capital structure channel of
monetary policy, Journal of Financial Economics 133, 357–378.
Hammersley, J. S., L. A. Myers, and C. Shakespeare, 2008, Market reactions to the
disclosure of internal control weaknesses and to the characteristics of those
weaknesses under section 302 of the Sarbanes Oxley Act of 2002, Review of
Accounting Studies 13, 141–165.
Hau, H., and S. Lai, 2016, Asset allocation and monetary policy: evidence from the
eurozone, Journal of Financial Economics 120, 309–329.
Henry, E., 2018, The information content of tax expense: a discount rate explanation,
Contemporary Accounting Research 35, 1917–1940.
Horton, J., 2007, The value relevance of ‘realistic reporting’: evidence from UK life
insurers, Accounting and Business Research 37, 175–197.

Horton, J., G. Serafeim, and I. Serafeim, 2013, Does mandatory IFRS adoption
improve the information environment?, Contemporary Accounting Research 30,
388–423.
Jacob, R. A., S. El-Gazzar, and S. McGregor, 2017, Evolution of financial reporting of
life insurers: the predominance of unregulated embedded value disclosure, Journal of
Financial Regulation and Compliance 25, 56–72.
Jiang, X., and B.-S. Lee, 2007, Stock returns, dividend yield, and book-to-market ratio,
Journal of Banking and Finance 31, 455–475.
Kim, Y., and M. S. Park, 2009, Market uncertainty and disclosure of internal control
deficiencies under the Sarbanes-Oxley Act, Journal of Accounting and Public Policy 28,
419–445.
Lambert, R., C. Leuz, and R. E. Verrecchia, 2007, Accounting information, disclosure,
and the cost of capital, Journal of Accounting Research 45, 385–420.
Lin, X., C. Wang, N. Wang, and J. Yang, 2018, Investment, Tobin’s q, and interest
rates, Journal of Financial Economics 130, 620–640.
Liu, D., H. Gu, and T. Xing, 2016, The meltdown of the Chinese equity market in the
summer of 2015, International Review of Economics and Finance 45, 504–517.
Maio, P., 2014, Another look at the stock return response to monetary policy actions,
Review of Finance 18, 321–371.
Malafronte, I., M. G. Starita, and J. Pereira, 2018, The effectiveness of risk disclosure
practices in the European insurance industry, Review of Accounting and Finance 17,
130–147.
Milian, J. A., 2015, Unsophisticated arbitrageurs and market efficiency: overreacting to
a history of underreaction?, Journal of Accounting Research 53, 175–220.
Milliman, 2017, 2016 Embedded Value Results: Europe. Available at: https://www.mil-
liman.com/-/media/milliman/importedfiles/uploadedfiles/insight/2017/2016-embed-
ded-value-europe.ashx
Oh, K., W. Choi, S. W. Jeong, and J. Pae, 2014, The effect of different levels of internal
control over financial reporting regulation on the quality of accounting information:
evidence from Korea, Asia-Pacific Journal of Accounting and Economics 21, 412–442.
Patelis, A. D., 1997, Stock return predictability and the role of monetary policy, The
Journal of Finance 52, 1951–1972.
Pham, H. N. A., V. Ramiah, and I. Moosa, 2020, The effects of environmental
regulation on the stock market: the French experience, Accounting and Finance 60,
3279–3304.
Prefontaine, J., J. Desrochers, and L. Godbout, 2009, The information content of
voluntary embedded value (EV) financial disclosures by Canadian life insurance
companies, International Business and Economics Research Journal 8, 1–16.
PwC (PricewaterhouseCoopers), 2007, Insurance reporting at the crossroads: what do
analysts think? Available at: https://pwc.blogs.com/files/global-ifrs-analysts-survey.
pdf
PwC (PricewaterhouseCoopers), 2009, Making sense of the numbers: analysts’ perspec-
tives on current and future reporting in the insurance industry. Available at: https://
www.pwc.com/gx/en/insurance/ifrs/assets/analyst-reporting-perspectives.pdf
Qi, B., L. Li, Q. Zhou, and J. Sun, 2017, Does internal control over financial reporting
really alleviate agency conflicts?, Accounting and Finance 57, 1101–1125.
Riedl, E. J., and G. Serafeim, 2011, Information risk and fair values: an examination of
equity betas, Journal of Accounting Research 49, 1083–1122.
Rogers, W., 1983, Analyzing complex survey data (Rand Corporation Memorandum,
Santa Monica, CA).

Serafeim, G., 2011, Consequences and institutional determinants of unregulated

corporate financial statements: evidence from embedded value reporting, Journal of
Accounting Research 49, 529–571.
Swiss Re, 2019, World insurance: the great pivot east continues. Available at: https://
www.swissre.com/dam/jcr:b8010432-3697-4a97-ad8b-6cb6c0aece33/sigma3_2019_en.
pdf
Thompson, S. B., 2011, Simple formulas for standard errors that cluster by both firm
and time, Journal of Financial Economics 99, 1–10.
Vedani, J., N. El Karoui, S. Loisel, and J.-L. Prigent, 2017, Market inconsistencies of
market-consistent European life insurance economic valuations: pitfalls and practical
solutions, European Actuarial Journal 7, 1–28.
Vuolteenaho, T., 2002, What drives firm-level stock returns?, The Journal of Finance 57,
233–264.
Wu, R. C., and A. W. Hsu, 2011, Value relevance of embedded value and IFRS 4
insurance contracts, The Geneva Papers on Risk and Insurance – Issues and Practice 36,
283–303.
Zimmermann, J., S. Veith, and J. Schymczyk, 2015, Measuring risk premiums using
financial reports and actuarial disclosures, The Geneva Papers on Risk and Insurance –
Issues and Practice 40, 209–231.

The Informativeness of Embedded Value Reporting To Stock Price

Uploaded by

Copyright:

Available Formats

The Informativeness of Embedded Value Reporting To Stock Price

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Informativeness of Embedded Value Reporting To Stock Price

Uploaded by

Copyright:

Available Formats

Accounting & Finance 61 (2021) 5341–5376

The informativeness of embedded value reporting to

Derrick W. H. Funga , David Joub, Ai Ju Shaoc, Jason J. H. Yehd

This paper examines the informativeness of embedded value reporting to stock

Key words: Embedded value accounting; Price to embedded value ratio;

JEL classification: M41, G22

Although price-to-book (P/B) and price-to-earnings (P/E) ratios are com-

Please address correspondence to Derrick W. H. Fung via email: [email protected]

© 2021 Accounting and Finance Association of Australia and New Zealand

© 2021 Accounting and Finance Association of Australia and New Zealand

© 2021 Accounting and Finance Association of Australia and New Zealand

5-year horizon), around 60 percent of the cross-sectional variation in current P/

© 2021 Accounting and Finance Association of Australia and New Zealand

sectional variations in P/EV ratios that can facilitate investment decision-

© 2021 Accounting and Finance Association of Australia and New Zealand

Embedded value (year-end of t)

Figure 1 Elements of embedded value. [Colour ﬁgure can be viewed at wileyonlinelibrary.com]

© 2021 Accounting and Finance Association of Australia and New Zealand

price beyond that of US GAAP accounting. Also, Zimmermann et al. (2015)

3. Data and methodology

As embedded values are not available in commonly used ﬁnancial databases,

© 2021 Accounting and Finance Association of Australia and New Zealand

Name of insurer Jurisdiction

AIA Group Ltd. Hong Kong

© 2021 Accounting and Finance Association of Australia and New Zealand

Variable Mean Std. dev Min 25th 50th 75th Max N

Panel A: Whole sample

Mean Mean annual growth Mean annual

Panel B: Individual insurers

AIA Group Ltd. 2010–2017 1.667 0.107 0.012 0.199 8

© 2021 Accounting and Finance Association of Australia and New Zealand

Mean Mean annual growth Mean annual

© 2021 Accounting and Finance Association of Australia and New Zealand

Mean Mean annual growth Mean annual

© 2021 Accounting and Finance Association of Australia and New Zealand

observations are available for empirical analysis. El-Gazzar et al. (2015)

where ΔPt+1 = Pt+1 − Pt and ΔEVt+1 = EVt+1 − EVt. Taking natural

© 2021 Accounting and Finance Association of Australia and New Zealand

To incorporate the dividend components, Equation (2) is re-written as

pevt ¼ pevtþ1 þ gevtþ1 þ dtþ1 rtþ1 (5)

Iterating Equation (5) forward by N − 1 periods gives:

© 2021 Accounting and Finance Association of Australia and New Zealand

varðpevt Þ ¼ covðpevt ,pevt Þ

Dividing both sides of Equation (7) by varðpevt Þ, we have:

βt ðpev,NÞ þ βt ðgev,NÞ þ βt ðd, NÞ þ βt ðr,NÞ ¼ 1 (9)

© 2021 Accounting and Finance Association of Australia and New Zealand

pev j,tþN ¼ βt ðpev,NÞpev j,t þ ɛ j,t ðpev,NÞ

4.1. Baseline results

© 2021 Accounting and Finance Association of Australia and New Zealand

Projected horizon Firm-year (1) (2) (3) (4) Sum of

1 2002–2016 163 85.45% 4.85% –0.01% 9.71% 1

© 2021 Accounting and Finance Association of Australia and New Zealand

Proporon of variaon aributable to future price to embedded value rao

© 2021 Accounting and Finance Association of Australia and New Zealand

4.2. Decomposing persistency in P/EV ratios

© 2021 Accounting and Finance Association of Australia and New Zealand

Dividing both sides of Equation (12) by varðpevt Þ, we have:

© 2021 Accounting and Finance Association of Australia and New Zealand

(11.23)* (2.94)* (0.00) (1.05)

(5.00)* (3.01)* (−1.16) (0.47)