The Quarterly Review of Economics and Finance 72 (2019) 159–167

Contents lists available at ScienceDirect

The Quarterly Review of Economics and Finance

journal homepage: www.elsevier.com/locate/qref

Comparing relative valuation efficiency between two stock markets

Ronghua Yi a , Yu-Wei Chang a,∗ , Wen Xing b , Jun Chen c
a
China Jiliang University, Xueyuan RD., NO. 258, Xiasha High Education Zone, Hangzhou, Zhejiang Province, PR China
b
Hong Kong Baptist University, Kowloon Tong Kowloon, Hong Kong
c
Auckland University of Technology, 55 Wellesley Street East, Auckland Central, New Zealand

a r t i c l e i n f o a b s t r a c t

Article history: How to compare the valuation efficiency between two stock markets? In the perspective of input-output,
Received 16 April 2018 relative valuation and multi-factors, a DEA model that can be used to evaluate the relative valuation
Received in revised form 21 July 2018 efficiency of two stock markets based oncross-listed stock samples was constructed. A DEA-Malmquist
Accepted 24 November 2018
model was also constructed to analyze the their valuation efficiency changes. We analyzed the valuation
Available online 27 November 2018
efficiency of mainland China and Hong Kong stock markets based on A+H cross-listed stocks from 2002
to 2013. The results were compared using the model we constructed andusing the P/E ratio. It shows that
Keywords:
as an emerging stock market, mainland stock market is still immature and needs more development.
Data envelopment analysis
Relative valuation efficiency
There is a linkage between the two stock markets and their valuation efficiency is influenced by similar
A+H cross-listed stocks environmental factors. The government intervention in mainland stock marketcan effectively improve
Comprehensive valuation factors market relative valuation efficiency. Compared with the P/E ratio, the DEA model is more comprehensive
Malmquist index and can match the actual circumstances better.
© 2018 Board of Trustees of the University of Illinois. Published by Elsevier Inc. All rights reserved.

1. Introduction 25 and 40. However, Allen and Gorton (1993) stated that stock
price bubbles originate from spontaneous behavior in the market.
Market efficiency is an important index for measuring the Damodaran (2001) stated that, in addition to the fundamentals of
investment value of a stock market. According to Tobin (1984), mar- a listed company itself, a reasonable P/E ratio is related to inter-
ket efficiency can be divided into different types, and an important est rates, the growth rate of GDP, and particular risk factors within
type is fundamental valuation efficiency. Fundamental valuation a country or region. In practical application, because of the limi-
efficiency describes the degree to which the price of a stock can tations of the P/E ratio (i.e., it is the only ratio of a stock price to
reflect its real economic value. Excessive valuation and insufficient its current earnings per share) and the difficulty in determining a
valuation are both regarded as inefficient. For each stock, its real reasonable standard for it (i.e., it is affected by factors, not only fun-
economic value is difficult to measure. Because of the complex- damentals), it is difficult to explain the large difference of average
ity of stock markets, stock prices and valuation efficiency can be P/E ratios among various stock markets in the long term (e.g., the
easily influenced by market interest rates, substitute investment New York and Tokyo stock markets are both mature stock mar-
tools and many other factors. In such circumstances, the aver- kets, but a large difference exists between their average P/E ratios)
age price/earnings (P/E) ratio index is generally regarded as the and to explain the large difference in the same stock market in
substitute indicator for measuring the pricing level and valuation various periods (e.g., the fluctuation amplitude of the average P/E
efficiency of stock markets. ratio of the New York S&P 500 was as high as eight times in the
A reasonable P/E ratio in a mature stock market should be past 90 years). Explaining the difference between the P/E ratios of
between 10 and 20, and each stock has little difference. Graham cross-listed stocks in various markets is also difficult.
and Dodd (1934), the founder of securities analysis, showed that a To solve the existing problems regarding the P/E ratio and other
reasonable P/E ratio is relevant to market maturity. The P/E ratio traditional test methods of market valuation efficiency, Campbell,
of stocks in a mature stock market should be approximately 15, Lo, and MacKinlay, (1997) stated that the primary disadvantage of
and that of high-growth stocks should be higher, maybe between traditional test methods is that they focus on whether a stock mar-
ket is absolutely efficient. The existence of market liquidity implies
the existence of profit opportunities. Absolute efficiency is an ideal
situation that will never appear. Hence, they introduced the con-
∗ Corresponding author at: Xueyuan RD., NO. 258, Xiasha High Education Zone,
cept of relative valuation efficiency, which is the efficiency of stocks
Hangzhou, Zhejiang Province, 310018, PR China.
E-mail addresses: [email protected] (R. Yi), [email protected] (Y.-W. Chang), relative to another one and is thought to be more useful than the
[email protected] (W. Xing), [email protected] (J. Chen). traditional absolute valuation efficiency.

https://doi.org/10.1016/j.qref.2018.11.008
1062-9769/© 2018 Board of Trustees of the University of Illinois. Published by Elsevier Inc. All rights reserved.
160 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167

Several scholars have explored the measurement of market 2. Research hypothesis and theoretical model
valuation efficiency from relative and multi-factor perspectives.
Li (2002) constructed a stochastic production frontier valuation 2.1. Research hypothesis
model by using a national macroeconomic and financial charac-
ter index. For each stock, he calculated the distance of its efficiency Pricing is the core function of stock markets, and the economic
deviation from the market frontier and viewed this as the mea- value of a stock is the benchmark of its pricing. However, the exis-
surement of its valuation inefficiency. Chan, Wu, and Kwok, (2007) tence of the size premium (Aziz & Ansari, 2014; Baetge, Kirsch,
proposed a model to measure relative valuation efficiency of stocks Koelen, & Schulz, 2010; Gharghori, Hamzah, & Veeraraghavan,
by comparing relative indices based on the stochastic frontier 2010), risk premium (Lam & Li, 2014), and liquidity premium
method; achieving high relative valuation efficiency indicates that, (Zhang & Pereira, 2008; Pasquierdoumer, 2013; Shen, 1993) indi-
compared with other stocks, a stock is able to achieve front pricing. cates that factors, not only fundamentals, may influence the stock
Cajueiro and Tabak (2004) proposed that the Hurst index can be cal- pricing and its valuation efficiency (Allen & Yang, 2004; Aretz et al.,
culated using rolling samples; the total time that a market deviates 2010; Harvey, 1994). In addition, stock pricing is the result of inter-
from the efficient situation can be added to evaluate the relative action and intercomparison in the stock market, so the pricing
valuation efficiency of the market. Ma (2004) proposed a method mechanism is complex and time varying. It is impossible to valu-
to calculate the relative valuation efficiency by using the correla- ate stock price accurately. In such circumstances, researching stock
tion coefficient of the daily return series.LLim, Hinich, and Brooks, pricing and market valuation efficiency from a comprehensive-
(2006) and Lim (2007) proposed that an autocorrelation coefficient valuation and relative-pricing perspective is reasonable, practical
can be used between the current return and early return to eval- and feasible.
uate the relative valuation efficiency through the use of a double In this study, to research the pricing mechanism and valua-
correlation test statistic (i.e., an H statistic) to test the existence tion efficiency of stock markets, we put forward four hypothesis
of nonzero double correlation. Evans (2006) proposed a method as follows.
to evaluate relative information efficiency by using the absolute
Hypothesis 1. The values of cross-listed stocks are related to
deviation of the variance ratio. Silva, Matsushita, and Giglio, (2008)
where they are listed.
evaluated the valuation efficiency based on the quantity of relative
information that was not translated efficiently into the stock price In the traditional theory, stocks, no matter where they are listed,
by computing the Lempel-Ziv complex index using the algorithmic be it the American stock market, European stock market, or Hong
complexity theory. Ito and Sugiyama (2009) used a time-varying Kong stock market, all have the same value, which means the val-
Auto-Regressive (AR) model to calculate the autocorrelation coef- uations of cross-listed stocks are exactly the same. However, when
ficient of a stock return in which the estimated AR coefficient of a listing in different stock markets, the interest rate, fund opportu-
Kalman filter changed with time; it was then viewed as the measure nity cost, investment channels, replacement of investment tools
of market inefficiency. Aretz, Bartram, and Pope, (2010) researched and many other factors are different, so the values of the same stock
the relation between the pricing of macroeconomic factors and in different stock markets are not all the same and are related to
market relative valuation efficiency by using multiple regressions the market they are listed in (King & Segal, 2008).
of stock price fluctuations and a set of macroeconomic factors (i.e.,
Hypothesis 2. In a valuation efficient stock market, the value of
economic growth expectations, inflation rate, total survival rate,
each stock is fully reflected by its price, so the valuation efficiency
and the term structure of interest rates and exchange rates).
of each stock and the entire stock market are equal to one.
Based on the literature analysis, this study investigates the rel-
ative valuation efficiency of stock markets from multi-factor and According to the definitions of the efficient market hypothe-
input-output perspectives. Stock markets are regarded as input and sis (EMH) and valuation efficiency, respectively proposed by Fama
output transcription systems. The relative valuation efficiencies (1970) and Tobin (1984), in a valuation efficient stock market, each
of different stock markets are compared by constructing a non- stock is priced reasonably (i.e., too high and too low are not ratio-
parametric production function (i.e., DEA) among the stock price, nal). In other words, in a valuation efficient stock market, if the
cumulative return rate and some related valuation factors. The valuation efficiency is viewed as the index to measure whether the
stock price and cumulative return rate are output variables, and price of each stock reflects its value, their valuation efficiencies are
other multi-factors related to stock pricing and valuation efficiency all equal to one, and their efficiency deviations are equal to zero.
are input variables. Thus, the relative and absolute valuation effi-
Hypothesis 3. The valuation efficiency of stocks cannot be fully
ciencies are both taken into consideration. The relative valuation
reflected by the rate of return.
efficiency in this study is measured in two dimensions: the average
DEA efficiency of all stocks in a market and their differences. More- Because the rate of return has better statistical characteristics
over, it is noticed that, in the real world, cross-listed stocks almost than price, it is widely used to evaluate the valuation efficiency of
always have different prices in different stock markets, which to stocks by calculating their cumulative excess returns. However, the
some degree reflects the valuation efficiency distinction of differ- expected return has time-varying characteristics, and the relations
ent stock markets. Thus, cross-listed stocks are used in this study as of price and rate of return are nonlinear. Thus, rate of return cannot
samples to compare the relative valuation efficiencies of different fully reflect stock price changes, and excess return is not a complete
stock markets. reflection of valuation premium. In such a field, only researching the
This study is organized as follows. Section 2 introduces the rate of return and ignoring the price is not appropriate. Combining
research hypothesis and theoretical model. Section 3 presents the the stock price and cumulative return rate is a good way to cover
empirical analysis, and we calculate the relative valuation efficien- the shortage.
cies of the Mainland China and Hong Kong stock markets based
Hypothesis 4. The absolute and relative pricing mechanisms of
on A+H cross-listed stock samples using both the DEA method and
stock markets can be better reflected by models with multi-factors.
the P/E ratio. The Malmquist indices of the two stock markets are
also calculated in this part. Section 4 provides the conclusions and Asset pricing is the core function of stock markets. A stock is a
suggestions for further study. type of special commodity whose investment value is not deter-
mined by its economic value alone. The existence of multi-factor
pricing mechanisms is proved by the three-factor model proposed
 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167 161

by Fama and French (1993). The complexity of stock markets and ⎧ N

the limitation of human cognitive ability determine that investors ⎪
⎪ 
⎪
⎪ xsji sj + Si− = sk xski
must reduce the dimensions of stock markets when making invest- ⎪
⎪
ment decisions. In this case, a relative valuation mechanism based ⎪
⎪ j=1;j =/ k
⎪
⎨  N
on multiple factors, including economic value and many other core
factors, is a good way to measure the pricing mechanism and valu- s.t. ysj sj − Si+ = ysk (2)
⎪
⎪
ation efficiency of stock markets. ⎪
⎪ j=1;j =/ k
⎪
⎪
⎪
⎪  sj ≥ 0; Si− ≥ 0; Si+ ≥ 0
⎪
⎩
i = 1, 2, ..., M; s = c, h
2.2. Theoretical model
According to the research hypothesis, the relative valuation effi-
2.2.1. Relative valuation efficiency of stock markets based on the ciency and the efficiency deviation of stock markets calculated by
DEA model the DEA model reflect the valuation efficiency of the entire stock
The DEA model, proposed by Charnes, Cooper, and Rhodes, market. For a valuation efficient stock market,  sk of all stocks
(1978), is a useful nonparametric model used to evaluate the rel- should equal one, and their efficiency deviation should be zero.
ative valuation efficiency of decision-making units (DMUs) that Thus, we define two indices in this study based on the DEA model to
have identical input and output variables. The DEA model was first reflect the relative valuation efficiency of stock markets, which are
applied to evaluate the performance appraisal of fund portfolios by shown in formulas (3) and (4). N represents the number of stocks
Murthi, Choi, and Desai, (1997). Powers and McMullen (2000) used in the market.
it to select large capitalization stocks. Abad, Thore, and Laffarga, (1)Market relative valuation level (Ls )
(2004) proposed a two-stage DEA model, using the financial infor- 
n
mation of companies to evaluate the fundamental value of stocks. Ls = sk /N (3)
Edirisinghe and Zhang (2007) proposed a two-stage integrated DEA k=1
model based on dynamic financial analysis for selecting investment
portfolios based on the DEA model. Dia (2009) proposed a four- (2)Market relative valuation efficiency (Es )
stage DEA model, where stock valuation factors, risks of stocks, and

n
 
the preference of the decision maker were regarded as input vari- Es = 1 − sk − Ls  /N (4)
ables for selecting a financial asset portfolio. The results of these
k=1
studies show that, without providing the production function in
advance, the DEA model with valuation factors and risk factors Obviously, according to the definitions of DEA and the EMH, the
as input variables agrees with the relationship of market valua- stock market is relative valuation efficient only when Ls and Es both
tion function that has complexity and time-varying characteristics equal one. In other words, if stock market s is efficient, Ls = Es = 1.
and can also consider both the absolute value and relative value of However, if stock market s is inefficient, not all  sk are equal to one,
stocks. and Ls < 1, Es < 1.
Thus, according to the modeling idea of DEA, a stock market
can be regarded as a complex input and output system, where the 2.2.2. DEA-Malmquist model
output variables are stock price and cumulative return rate and The Malmquist index is an important index for measuring the
the input variables are company fundamentals, macroeconomic relative productivity change of DMUs in multiple time periods by
factors, market factors, and investor preferences. Stock price and DEA. It can be broken down into two parts: efficiency change (EC)
cumulative return rate are the comprehensive expression of stocks’ and technological change (TC). The former shows the management
absolute value (i.e., economic value) and relative value. and organization efficiency in the market, and the latter is the
For cross-listed stocks, suppose that N (k = 1, 2,. . ., N) stocks consequence of innovation and adoption of new technologies. The
simultaneously exist in L (s = 1,2,. . ., L) stock markets; this produces efficiency change (EC) can also be broken down into pure efficiency
N × L DMUs. The output variables are stock price and cumulative change and scale efficiency change.
return rate expressed by ysk , and the input variables are M (i = 1, For stock market s, the definitions of the Malmquist index based
2,. . ., M) valuation factors expressed by xski . The relative valuation on the optimal efficiency frontier in t and (t + 1) years are shown in
efficiency of stock k in market s calculated by the DEA model is formulas (5) and (6), respectively.
shown in Formula (1), in which  sk (no more than one) is the DEA
relative valuation efficiency and ε is an Archimedes infinitesimal.
t
MIsk t
= sk (xjt , yjt )/sk
t
(xjt+1 , yjt+1 ) (5)

t+1 t+1 t t+1 t+1

MIsk = sk (xj , yjt )/sk (xj , yjt+1 ) (6)
M
max sk = ysk / v x
i−1 si ski t (xt , yt ) is the relative valuation efficiency of
 M In formula (5), sk j j
ysk / v x ≤ 1, k = 1, . . ., N, i = 1, . . .M (1) t (xt+1 , yt+1 )
i−1 si ski stock market s in year t based on the dataset in year t. sk
s.t. j j
vsi ≥ ε is the relative valuation efficiency of stock market s in year (t + 1)
t+1 t
based on the dataset in year t. In formula (6), sk (xj , yjt ) is the
relative valuation efficiency of stock market s in year t based on
Using the Charnes-Cooper transformation and linear programming the dataset in year (t + 1). sk t+1 t+1
(xj , yjt+1 ) is the relative valuation
dual theory and introducing the S+ and S− slack variables, the
efficiency of stock market s in year (t + 1) based on the dataset in
dual programming model without Archimedes infinitesimals is as
year (t + 1).
shown in Formula (2).
To avoid calculating the Malmquist index by choosing an
optimal efficiency frontier arbitrarily, we applied the geomet-
 M 
 ric mean method to calculate it, just as shown in formula (7).
Min sk − ε Si− + Si+ MI(xjt+1 , yjt+1 ; xjt , yjt ) is the efficiency change from year t to year
i=1 (t + 1).
162 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167

Zhang (2007) identified a total of 18 financial parameters from

MI(xjt+1 , yjt+1 ; xjt , yjt ) = t
{[sk (xjt , yjt )/sk
t
(xjt+1 , yjt+1 )] × t+1 t
[sk t+1 t+1
(xj , yjt )/sk (xj , yjt+1 )]}
1/2 financial statements of companies. These parameters examine a
firm’s fundamental performance through a range of performance
(7) perspectives: profitability, asset utilization, liquidity, leverage, val-
uation, and growth perspectives. Powers (2000) chose 5 output
In accordance with the theory of the Malmquist index, it can be variables and 3 input variables from the perspective of stock market
broken down into efficiency change (EC) and technological change performance and risk.
(TC), which are shown in formulas (8) and (9), respectively. ECsk >1, Because the variables in the DEA model are selected based on
ECsk <1 and ECsk = 1 represent that, from year t to year (t + 1), the different application purposes, input and output variables are very
efficiency appears to increase, decrease and remain unchanged, flexible. The variables included depend on the application and
respectively. Similarly, TCsk >1, TCsk <1 and TCsk = 1 represent that, research issues. The diversity can be found in Table 1.
from year t to year (t + 1), there is a positive technological change, According to stock pricing theory and the discussion in Section
negative technological change and no technological change, respec- 2, refer to the previous studies on DEA model variables. The out-
tively. The corresponding indices of the entire market are the put variables should reflect the current stock price and its changes.
arithmetic mean value of all stock samples in the market. Thus, stock price and cumulative return rate were selected as the
output variables in the DEA model. For the two output variables,
t+1 t+1
t
ECsk = sk (xjt , yjt )/sk (xj , yjt+1 ) (8) we chose the closing date of the first trading week in May as the
time point every year. We use the closing price of that day and the
t+1 t+1
TCsk = {[sk (xj , yjt+1 )/sk
t
(xjt+1 , yjt+1 )] total rate of return from the first day of the year to that day as the
stock price variable and the cumulative return rate variable of the
t+1 t
×[sk (xj , yjt )/sk
t
(xjt , yjt )}
1/2
(9) last year, respectively (e.g., we use the closing price of May 3rd 2003
as the stock price variable of the 2002 DEA model and add the rate
of return from January 1st to May 3rd 2003 as the cumulative return
rate variable of the 2002 DEA model)1 .
3. Empirical analysis According to previous studies (Chen, 2008; Serifsoy, 2007;
Tsolas, 2013; Wilkens & Zhu, 2001), we select five input variables
The data of A+H cross-listed stock samples were entered into that can better reflect the fundamental and technical indicators.
Formula (2) to construct the DEA models of the Mainland China The fundamental indicators include the intrinsic value per share,
and Hong Kong stock markets using the DEAP2.1 software. On the weighted average growth rate in the main business of each com-
one hand, different stock markets have their own valuation modes, pany, proportion of cash dividends per share, and coefficient of the
so the DEA models of the two stock markets should be constructed circulation scale, while the technical indicator includes the risks of
independently. On the other hand, to compare the two stock mar- stocks. The specific definitions of each variable and their calculation
kets on the same standard, all stock samples should be used to methods are shown as follows.
construct a unified DEA model. In this study, we first constructed
different DEA models for each stock market. Then, we constructed
a unified DEA model using all stock samples, calculating and com-
paring their Ls and Es indices to perform a better comparison. Just as 3.1.1.1. Intrinsic value per share. According to the correction F–O
mentioned above, the P/E ratio is regarded as the substitute indica- model proposed by Liu [28], investors can only predict the devel-
tor for measuring the pricing level and valuation efficiency of stock opment of companies for a limited N years. The model is shown in
markets. After analyzing the results of the DEA models, we calcu- Formula (10).
lated the results of the P/E ratio using the same stock samples as
the DEA models to compare the two methods (i.e., the DEA method
and the P/E ratio method) in measuring the valuation efficiency of 
N
ROE − 
stock markets. Pjt (FV ) = BVjt + × (1 + ROE − ˛ × ROE − )i−1 × BVjt
(1 + )i
i=1
3.1. Variable and sample selection (10)

3.1.1. Variable selection

The choice of DEA model variables should be based on differ-
ent application purposes, including the main input and output In Formula (10), ROE, 2 , and ˛respectively represent the aver-
variables. The input variables can be generally divided into two age rate of return on equity, average rate of risk-free interest, and
types: technical and fundamental indicators (Murphy, 1999). Fun- average dividend ratio.3 In this study, N was 10 years.
damental indicators tend to be associated with trading, such as
a company’s management structure, competitors, industry posi-
tion, growth rate, growth potential, income, and revenues. The
latter tend to be associated with investor behavior, such as patterns 1
The reason why we chose the closing date of the first trading week in May as
within stock charts. the time point every year is that, in the Mainland China stock market, the deadline
There have been some research studies that have applied DEA for a company to publish its financial statement is April 30th of the following year.
A period of time is required for stock markets to review the information published
models in mutual fund performance evaluation and the most desir-
in financial statements and to eliminate the influence of short-term fluctuations.
able securities selection. Murthi et al. (1997) put forward a portfolio 2
The data for are the closing rate of 5-year government bonds in the first week
performance measurement based on the DEA portfolio efficiency of May.
index (DPEI) with risk and cost as inputs and excess return as out- 3
When calculating ROE, , and ˛, the weights of data in years t, (t -1), and (t -2)
put. Zhao and Wang (2007) selected the unit net value and term are used to calculate that their average indices are 0.6, 0.3, and 0.1, respectively. If t
is the first year that a company is on the market, the data of t itself are the average
return of the net value as outputs and the comparative standard data; if (t -1) is the first year, the weights of data in years t and (t -1) are 0.7 and 0.3,
deviation, percentage of negative monthly return and operation respectively. The methods of calculating the weighted indexes in (b) and (c) are the
fee as inputs. For the issue of stock selection, Edirisinghe and same.
 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167 163

Table 1
Variables selected in previous studies.

Input variable
Previous studies Output variable Research issue
Fundamental indicators Technical indicators

Chen (2008) revenues, operating profit, net Average equity, average asset, Stock selection to construct
income sales cost portfolios
Edirisinghe and Zhang (2007) Return on equity, Return on Receivables turnover, Predictive of firm’s stock price
assets, Net profit margin, Price Inventory turnover, Asset returns for equity portfolio
to earnings (PE) ratio, Price to turnover, Current ratio, Quick management
book ratio, Earnings per share ratio, Debt to equity ratio,
(EPS), Revenue growth rate, Leverage ratio, Solvency ratio-I,
Net income growth rate, Solvency ratio-II
Earnings per share growth rate
Murthi et al. (1997) Excess returns Expense ratio, Loads Standard deviation of efficiency of mutual funds and
return, Turnover portfolio performance of
measurement
Powers (2000) 1-Year Return, 3-Year Price to earnings ratio Beta (risk) (5 year), Sigma select the most desirable
Return,5-Year Return,10-Year (5 year) securities
Return, Earnings per share
Serifsoy (2007) Number of listed companies at Number of staff employed at Stock exchange business
exchange in period, Total exchange in period, Tangible models
trading volume in bonds and assets at exchange in period
shares at exchange in period,
Total number of derivatives
contracts traded at exchange in
period, Post-trading services
and software sales at exchange
in period
Tsolas (2013) total revenue, net income Net fixed assets, Total equity, Profitability and stock market
before taxes, and market Total operating cost, Selling performance of listed
capitalization. and administrative expenses, construction firms
Wilkens and Zhu (2001) Minimum return standard deviation of Evaluate the performance of
returns, proportion of turn, mutual funds
skewness of return
distribution
Zhao and Wang (2007) Unit net value, Term return of Operation fee Comparative standard Evaluate the performance of
net value deviation, Percentage of mutual funds
negative monthly return

3.1.1.2. Weighted average growth rate in the main business. To The five input variables of the DEA model selected in the study
reflect the growth trend of a company objectively, data in the most are all comprehensive variables, so even if there are only five input
recent 3 years were used to calculate the variable.4 variables, they could already fully reflect a company in various
aspects.
3.1.1.3. Proportion of cash dividends per share. Data from the most
recent 3 years were used to calculate the variable. 3.1.2. Sample selection and source of data
To compare the relative valuation efficiency of the Mainland
3.1.1.4. Coefficient of circulation scale. Banz and many other China and Hong Kong stock markets, A+H cross-listed stocks that
researchers have shown the existence of the scale effect and stock existed in both two stock markets in the same year were selected
liquidity premium [7]. The variable was defined as the ratio of a as samples.5 The number of samples is listed in Table 2 (i.e., the
stock’s circulation scale to the market circulation scale. total number of A and H stock samples). The sources of data are
a wind database, world bank (http://data.worldbank.org.cn/) and
China National Statistics Yearbook (2003∼2014). The data analysis
3.1.1.5. Risks of stocks. Total risk can be divided into systematic risk software is DEAP2.1 and Eviews.
and non-systematic risk. The regression models used to evaluate
risks of stocks are shown in Formulas (11) and (12). In the formula,
ri is the return rate of stock i calculated by the data of weekly closing 3.2. Analysis based on the results of the DEA model
prices. rI is the return rate of the stock market (i.e., in this study,
data from the Shanghai Composite Index and Hang Seng Index were 3.2.1. Analysis based on the results of different DEA models
selected to calculate the market return rate) in the same period as Under the view that the valuation modes of different stock mar-
ri .  i 2 is the total risk of stock i;  I 2 is the variance of market risks; kets are not all the same, in this part, the DEA models of the two
ˇi 2  I 2 is the systematic risk of stock i; and  εi 2 is the variance of stock markets (i.e., the mainland stock market and Hong Kong stock
εi , which represents the non-systematic risk of stock i. market) were constructed separately. The DEA relative valuation
efficiency of each stock was calculated. Based on the DEA efficiency
ri = ˛i + ˇi rI + εi (11) of each stock, the Ls and Es indices of the two stock markets were
also calculated, respectively, according to Formulas (3) and (4). The
i2 = V (ri = ˛i + ˇi rI + εi ) = ˇi2 I2 + εi
2
(12) results are shown in Figs. 1 and 2.

4 5
According to the DEA theory, input variables should be positive, and the non- Because the number of A+H cross-listed stocks was small before 2002, stocks
negative treatment did not affect the results. Thus, to avoid negative input variables, that existed in both stock markets in the period from 2002 to 2013 were selected as
“1” is added to each weighted index of the growth rate. empirical analysis samples. The number of samples increased each year.
164 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167

Table 2
Number of A+H cross-listed samples.

year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013

Number of samples 52 54 56 58 64 88 106 110 118 140 150 164

Fig. 1. Comparison of Ls in different DEA models. Fig. 3. Comparison of Ls in the same DEA model.

Fig. 4. Comparison of Es in the same DEA model.

Fig. 2. Comparison of Es in different DEA models.

relative valuation efficiency. The price limit in the mainland stock

Fig. 1 shows the relative valuation level Ls (s=c,h) of the two market is also helpful.
stock markets from 2002 to 2013 based on different DEA mod-
els. We can see that the curve of the Lh index in the Hong Kong 3.2.2. Analysis based on the results of the same DEA model
stock market has less fluctuation than that of the Lc index in the Considering that each stock market has its own valuation mode,
mainland stock market, indicating that the Hong Kong stock market we constructed different DEA models for the mainland and Hong
has relatively better stability in stock pricing and is more mature. Kong stock markets in 3.2.1. In this part, to compare the two stock
The trends and fluctuation directions of the two stock markets are markets in the same pricing standard, all A+H cross-listed stock
nearly the same except for several years. The different fluctuation samples were used to construct a unified DEA model. The results
direction of the two stock markets around 2008 may be relevant to are shown in Figs. 3 and 4.
the global financial crisis in 2008. In Fig. 3, we can see that the curve of the Lh index in the Hong
Fig. 2 shows the relative valuation efficiency Es (s=c,h) of the two Kong stock market is lower than that of the Lc index in the main-
stock markets from 2002 to 2013 based on different DEA models. land stock market, which is different from the result in Fig. 1. It
First, we can see that the curve of the Eh index in the Hong Kong indicates that, when comparing all A+H cross-listed samples in the
stock market is more stable than that of the Ec index in the main- same pricing standard, the relative valuation level of the mainland
land stock market, indicating that the Hong Kong stock market is stock market is generally higher than that of the Hong Kong stock
more mature, which is similar to the result in Fig. 1. Second, Fig. 2 market, which is consistent with the actual situation that stocks in
shows that the relative valuation efficiency in the mainland stock the mainland stock market have a higher pricing level. In addition,
market is higher than that of the Hong Kong stock market in most we notice that the fluctuation directions of Ls (s=c,h) of the two
years. Considering that the market intervention is greater in the stock markets are different. A probable reason for that may be the
mainland stock market, the results shown in Fig. 2 indicate that, trend of Lc in the mainland stock market is relatively independent
although the existence of government power and market inter- and is already out of line with the fundamentals to some degree.
vention is a double-edged sword, for now, it does more good than The results shown in Fig. 4 are similar to those in Fig. 2. The
harm. It restrains the excessive volatility of stock prices, avoids irra- fluctuation of Eh in the Hong Kong stock market is smaller than
tional investor overreaction and can effectively improve the market that of Ec in the mainland stock market, and the relative valuation
 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167 165

Fig. 5. Comparison of Ls based on P/E ratio.

Fig. 7. Comparison of Ls based on P/E and DEA.

Fig. 6. Comparison of Es based on P/E ratio.

Fig. 8. Comparison of Es based on P/E and DEA.

efficiency in the mainland stock market is higher than that of the

Hong Kong stock market in most years. 3.3.2. Comparison of the DEA and the P/E ratio methods
To compare the DEA method and the P/E ratio method in mea-
3.3. Comparison of the DEA and the P/E ratio methods suring the valuation level and valuation efficiency of stock markets,
the Lc /Lh and Ec /Eh indices based on the two different methods were
3.3.1. Analysis based on the results of the P/E ratio index selected for comparison. The results are shown in Figs. 7 and 8.
We calculated the results of the P/E ratio in both of the two Fig. 7 shows the ratio of the valuation level (Lc /Lh ) from 2002
stock markets. The samples and the calculation method of the Ls to 2013 based on the two different methods. First, the curves of
and Es indices are the same as those of the DEA model according Lc /Lh based on the two methods are all greater than one, indicating
to formulas (3) and (4). As in the DEA model, we chose the closing that the valuation level in the mainland stock market is higher than
date of the first trading week in May as the time point every year. that in the Hong Kong stock market. Second, while the fluctuation
We used the closing price of that day and the earnings per share trends of the two curves are similar, the curve based on the P/E
published in the last year’s financial report to calculate the P/E ratio ratio is higher than that based on the DEA model, and its fluctuation
of last year. The Ls index is used to measure the valuation level of amplitude is wider. It indicates that the gap in the valuation level
the stock market, and the Es index is used to measure the valuation (Ls ) between the mainland and Hong Kong stock markets is larger
efficiency of the stock market. The results of the P/E ratio are shown when measured by the DEA method. Because the DEA model is
in Figs. 5 and 6. more comprehensive than the P/E ratio, it is evident that factors,
Fig. 5 shows the valuation levels Ls (s=c,h) of the mainland and not only profit, were priced, and the gap in Ls between the two stock
Hong Kong stock markets from 2002 to 2013 based on the P/E ratio markets could not be accurately reflected by the P/E ratio alone. The
index. The curve of the Lh index in the Hong Kong stock market is gap in the valuation level between the two stock markets shown
lower than that of the Lc index in the mainland stock market, which by the P/E ratio may be overestimated, while the gap shown by the
is similar to the results shown in Fig. 3 when using the DEA method. DEA model better fit the real situation. In addition, the differences
The curve of the Lh index in the Hong Kong stock market has less of the two curves in 2002 may be related to the small sample size.
fluctuation than that of the Lc index in the mainland stock market, Fig. 8 shows the ratio of the valuation efficiency (Ec /Eh ) from
indicating that the Hong Kong stock market is more mature, which 2002 to 2013 based on the two different methods. The trends of the
is also similar to the result in the DEA model. two curves are basically consistent. It seems that the curve based on
Fig. 6 shows the valuation efficiencies Es (s=c,h) of the main- the P/E ratio is approximately one year ahead of that based on the
land and Hong Kong stock markets from 2002 to 2013 based on the DEA model. In general, the P/E ratio only uses the data of stock price
P/E ratio. It seems that there is an upward trend in the Ec and Eh and earnings per share, while in the DEA model, many other factors
indices, and the fluctuations of the two indices are similar, which related to market valuation efficiency, such as cumulative return
is inconsistent with the result in the DEA model. rate, circulation scale, growth rate and stock risks, are taken into
166 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167

Table 3
Malmquist index of mainland and Hong Kong stock markets and its decomposition.

Mainland stock market (c) Hong Kong stock market (h)

period
Efficiency change (EC) Technological change (TC) Malmquist Index Efficiency change (EC) Technological change (TC) Malmquist Index

2002∼2003 0.88 1.38 1.22 0.71 1.55 1.09

2003∼2004 0.98 0.58 0.57 2.19 0.71 1.54
2004∼2005 0.96 1.18 1.14 0.76 1.31 0.99
2005∼2006 1.30 1.62 2.12 0.91 1.50 1.37
2006∼2007 0.78 0.78 0.61 1.37 0.71 0.97
2007∼2008 1.27 0.79 1.01 0.98 0.79 0.77
2008∼2009 0.87 1.26 1.11 1.08 1.33 1.44
2009∼2010 1.06 0.86 0.91 0.93 0.81 0.75
2010∼2011 1.01 0.91 0.92 0.94 0.93 0.88
2011∼2012 0.89 1.38 1.23 0.94 1.31 1.23
2012∼2013 1.03 0.75 0.77 1.04 0.77 0.80
average 1.00 1.05 1.05 1.08 1.07 1.08

consideration. When calculating the P/E ratio of the entire stock tion effectiveness and dynamic evolution of the mainland and Hong
market, the impact of samples that have negative indices must Kong markets. We further compare the analysis results between
be excluded. Thus, compared with the P/E ratio, the DEA method the DEA-Malmquist index and the traditional P/E ratio index. The
is more comprehensive, more reliable and can match the actual conclusions are as follows.
circumstances better. First, during the period between 2002 and 2013, although the
relative valuation level of the mainland market is higher than that
3.4. Analysis based on the results of the DEA-Malmquist model of the Hong Kong market, the relative valuation efficiency is lower
than that of the Hong Kong market. The gap between the two
According to the data of the A+H cross-listed samples and the markets continues to narrow, indicating that the valuation effec-
theoretical model of the DEA-Malmquist, the productivity changes tiveness of the mainland market is increasing. In addition, the
(Malmquist index), the efficiency changes (EC) and the technologi- fluctuation trends of the two markets are basically consistent, and
cal changes (TC) of the mainland and Hong Kong stock markets are the co-movement between the two markets is strong. The full cir-
as listed in Table 2. culation reform of the mainland market has a short-term negative
First, from Table 3, we can see that the fluctuation amplitudes impact on the valuation efficiency, and the market turbulence trig-
of the Malmquist index, the EC and the TC in the mainland stock gered by the global financial crisis has a long-term negative impact
market from 2002 to 2013 are greater than those in the Hong Kong on the valuation efficiency.
stock market, and the averages of the three indices are lower than Second, according to the comparison analysis between the DEA
those in the Hong Kong stock market, indicating that, as an emerg- model and P/E ratio, the relative valuation level of the mainland
ing stock market, the mainland stock market is still immature and market is higher than that of the Hong Kong market, and the fluctu-
requires more development. ation trends of the two markets are basically consistent. However,
Second, the changing directions of the Malmquist index in the based on the P/E ratio, the gap between relative valuation levels and
two stock markets are basically the same. In other words, the the fluctuation trends of the two markets is greater. The results
Malmquist indexes of the two stock markets in the same period are show that the relative valuation level of the mainland market is
both greater than or less than one except for the period from 2003 higher than that of the Hong Kong market, which is inconsistent
to 2005 and the period from 2007 to 2008. This shows that there with the fact that the Hong Kong market is more mature than the
is probably a linkage between the two stock markets or that their mainland market in the actual development stage. In a compre-
valuation efficiency is influenced by similar environmental factors. hensive view, because the valuation level and efficiency based on
The differences of the Malmquist index from 2003 to 2005 may the P / E ratio excludes the negative item index, the index may
related to the full circulation reform implemented in the mainland overestimate the gap between the two markets and consider more
stock market. The differences of the Malmquist index from 2007 to valuation factors. Therefore, the DEA model is more suitable for
2008 may be relevant to the global financial crisis. They both have measuring the relative valuation level and efficiency than the P/E
had bad influences on the stock markets. ratio.
Finally, the EC of the two stock markets are significantly differ- Third, in terms of the comprehensive technical efficiency
ent, while their TC are similar to each other, indicating that their changes, technological changes, and Malmquist index changes, the
technological changes are influenced by similar factors, but the fluctuation of the mainland market is greater than that of the Hong
operation and the management of the two stock markets are really Kong market, and the average value of the mainland market is
different. smaller than that of the Hong Kong market. This indicates that
the mainland market has immature characteristics. The changes
of the Malmquist index are basically the same each year, indicat-
4. Conclusions and prospects
ing that the efficiency changes of the two markets are influenced
by the common environmental factors, which in turn leads to a
This study is motivated based on three propositions: (1) all asset
strong co-movement between the two markets. The full circula-
prices in the efficient market reflect their real economic value, (2)
tion reform has a positive influence on the valuation efficiency of
the rate of return cannot fully reflect the valuation and charac-
the two markets. Because the financial crisis has a negative impact
teristics of the stock, and (3) the multi-factor model can better
on the valuation efficiency of the two markets, the market effec-
reflect the absolute and relative pricing mechanism. In the view
tiveness has not yet returned to the pre-crisis valuation.
of relative pricing, this study proposes a method for quantitatively
In summary, different from the other literature entries mainly
measuring the relative valuation level and efficiency of the mar-
concerned with relative Information-Arbitrage Efficiency (Cajueiro
ket based on the DEA multi-factor model. This study uses the A+H
& Tabak, 2004; Chan et al., 2007; Evans, 2006; Ito & Sugiyama, 2009;
cross-listed sample as the research object and examines the valua-
 R. Yi et al. / The Quarterly Review of Economics and Finance 72 (2019) 159–167 167

Ma, 2004; Silva et al., 2008), the research framework proposed in Edirisinghe, N. C. P., & Zhang, X. (2007). Generalized DEA model of fundamental
this study uses cross-listed stocks to realize the relative market analysis and its application to portfolio optimization. Journal of Banking &
Finance, 31(11), 3311–3335.
valuation level and the relative market valuation efficiency evalua- Evans, T. (2006). Efficiency tests of the UK financial futures markets and the impact
tion. This study integrates the efficient market hypothesis of Fama of electronic trading systems. Applied Financial Economics, 16(17), 1273–1283.
(1970), fundamental valuation efficiency concept of Tobin (1984), Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical
work. Journal of Financial Finance, 2(2), 383–417.
and relative efficiency thought of Campbell et al. (1997) to improve Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and
the valuation of securities, measurement of market efficiency, and bonds. Journal of Financial Economics, 33, 3–56.
comparative study of efficiency among markets. Gharghori, P., Hamzah, Y., & Veeraraghavan, M. (2010). Migration and its
contribution to the size and value premiums: Australian evidence. Journal of
There are still some limitations in the study. First, because DEA
International Financial Markets Institutions and Money, 20(2), 177–196.
is a nonparametric model, the stability test was difficult to per- Graham, B., & Dodd, D. L. (1934). Security analysis. McGraw Hill Financial
form. Second, the number of samples used in the empirical analysis Incorporation.
Harvey, C. R. (1994). Predictable risk and returns in emerging markets. Social
is insufficient, particularly in the first 4 years. It is worth noting
Science Electronic Publishing, 8(3), 773–816.
that because of the relativity characteristic of the DEA model, the Powers, J. (2000). Using data envelopment analysis to select efficient large market
relative valuation level and relative valuation efficiency of stock cap securities. Journal of Business & Management, 7(2), 31–42.
markets cannot reflect their absolute valuation efficiency. To reflect Ito, M., & Sugiyama, S. (2009). Measuring the degree of time varying market
inefficiency. Economics Letters, 103(1), 62–64.
the valuation efficiency of stock markets better, we should combine King, M. R., & Segal, D. (2008). Market segmentation and equity valuation:
the DEA model with the P/E ratio. The reliability of the results may Comparing Canada and the United States. Journal of International Financial
have been affected by these problems. However, these are areas Markets Institutions and Money, 18(3), 245–258.
Lam, K. S., & Li, F. K. (2014). The risk premiums of the four-factor asset pricing
that we can improve in future studies. As a preliminary study, the model in the Hong Kong Stock Market. Applied Financial Economics, 18(20),
results obtained in this study achieved the proposed objective. 1667–1680.
Li, K. (2002). What explains the growth of global equity markets? Canadian
Investment Review, 15(3), 23–30.
Acknowledgements Lim, K. P., Hinich, M., & Brooks, R. D. (2006). Events that shook the market: An
insight from nonlinear serial dependencies in intraday returns. In SSRN
This research was supported by the National Natural Sci- working paper.
Lim, K. P. (2007). Ranking of efficiency for stock markets: A nonlinear perspective.
ence Foundation of China (No. 71473235, No. 71871205, and No. Physica A-Statistical Mechanics and Its Applications, 376, 445–454.
71601172). Liu, H. (2005). The theory of the stock intrinsic value and the bubble of China’s
stock market. Economic Research Journal, 2, 45–53.
Ma, S. (2004). The efficiency of China’s stock market. Hampshire, UK: Ashgate.
References Murphy, J. J. (1999). Technical analysis of the financial Markets.pDf. Pennsylvania
Dental Journal, 77, 33–34.
Abad, C., Thore, S. A., & Laffarga, J. (2004). Fundamental analysis of stocks by Murthi, B. P. S., Choi, Y. K., & Desai, P. (1997). Efficiency of mutual funds and
two-stage DEA. Managerial and Decision Economics, 25(5), 231–241. portfolio performance measurement: A non-parametric approach. European
Allen, D. E., & Yang, W. (2004). Do UK stock prices deviate from fundamentals? Journal of Operational Research, 98(2), 408–418.
Mathematics and Computers in Simulation, 64(3–4), 373–383. Pasquierdoumer, L. (2013). Intergenerational transmission of self-employed status
Allen, F., & Gorton, G. (1993). Churning bubbles. The Review of Economic Studies, in the informal sector: A constrained choice or better income prospects?
60(4), 813–836. Evidence from seven west african countries. Journal of African Economies, 22(1),
Aretz, K., Bartram, S. M., & Pope, P. F. (2010). Macroeconomic risks and 73–111.
characteristic-based factor models. SSRN Electronic Journal, 34(6), 1383–1399. Powers, J., & McMullen, P. R. (2000). Using data envelopment analysis to select
Aziz, T., & Ansari, V. A. (2014). Size and value premiums in the indian stock market. efficient large market cap securities. Journal of Business & Management, 7(2),
Social Science Electronic Publishing, 7(4), 74–80. 31–42.
Baetge, J., Kirsch, H. J., Koelen, P., & Schulz, R. (2010). On the myth of size premiums Serifsoy, B. (2007). Stock exchange business models and their operative
in corporate valuation: Some empirical evidence from the german stock market. performance. Journal of Banking & Finance, 31(10), 2978–3012.
Social Science Electronic Publishing. Shen, P. (1993). Pricing bid-ask spreads in common stocks, liquidity premium and
Banz, R. W. (1981). The relationship between return and market value of common the small firm effect. General Information Programme, Unisist Newsletter.
stocks. Journal of Financial Economics, 9(1), 3–18. Silva, S. D., Matsushita, R., & Giglio, R. (2008). The relative efficiency of stock
Cajueiro, D. O., & Tabak, B. M. (2004). Ranking efficiency for emerging markets. markets. Economics Bulletin, 7(6), 1–12.
Chaos, Solitons, and Fractals, 22(2), 349–352. Tobin, J. (1984). . pp. 1–15. On the efficiency of the financial system. Fred hirsch
Campbell, J. Y., Lo, A. W., & MacKinlay, A. C. (1997). The econometrics of financial memorial lecture (153) New York: Lloyds Bank Review.
markets. New Jersey, US: Princeton University Press. Tsolas, I. E. (2013). Modeling profitability and stock market performance of listed
Chan, Y. C., Wu, C., & Kwok, C. (2007). Valuation of global IPOs: A stochastic frontier construction firms on the Athens exchange: Two-stage DEA approach. Journal
approach. Review of Quantitative Finance and Accounting, 29(3), 267–284. of Construction Engineering and Management, 139(1), 111–119.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of Wilkens, K., & Zhu, J. (2001). Portfolio evaluation and benchmark selection: A
decision making units. European Journal of Operational Research, 2(6), 429–444. mathematical programming approach. The Journal of Alternative Investments,
Chen, H. H. (2008). Stock selection using data envelopment analysis. Industrial (Summer), 9–19.
Management & Data Systems, 108(9), 1255–1268. Zhang, H. H., & Pereira, J. P. (2008). Stock returns and the volatility of liquidity. The
Damodaran, A. (2001). Investment valuation: Tools and techniques for determining Journal of Financial and Quantitative Analysis, 45(04), 1077–1110.
the value of any asset. New Jersey, US: John Wiley & Sons Press. Zhao, X.-j., & Wang, S.-y. (2007). Empirical study on chinese mutual funds’
Dia, M. (2009). A portfolio selection methodology based on data envelopment performance. Systems Engineering - Theory & Practice, 27(3), 1–11.
analysis. European Journal of Information Systems : an Official Journal of the
Operational Research Society, 47(1), 71–79.