Level of efficiency in the UK equity market: empirical study of the …ranadevajayasekera.net ›...
Transcript of Level of efficiency in the UK equity market: empirical study of the …ranadevajayasekera.net ›...
ORI GINAL RESEARCH
Level of efficiency in the UK equity market: empiricalstudy of the effects of the global financial crisis
Taufiq Choudhry • Ranadeva Jayasekera
� Springer Science+Business Media New York 2013
Abstract This paper investigates the effect of good or bad news (the asymmetric effect)
on the time-varying beta of firms in the UK during good periods (booms) and bad periods
(recessions). Daily data from twenty five UK firms of different sizes and from different
industries are applied in the empirical tests. The data ranges from 2004 to 2010, which
includes the current global financial crisis. The time-varying betas are created by means of
the bivariate BEKK GARCH model, and then linear regressions are applied to test for the
asymmetric effect of news on the beta. The asymmetric effects are investigated based on
both market and non-market shocks. Most firms and industries seem to support the market
efficiency hypothesis during both periods. However, the level of market efficiency seems to
decline significantly from the pre-crisis to crisis period. Both the results of market effi-
ciency and declining market efficiency from the pre-crisis to crisis periods provide ample
evidence of the asymmetric effect of the financial crisis on the beta of UK firms.
Keywords Asymmetric effect � Time-varying beta � BEKK � Market
efficiency � Asset mispricing
JEL Classification G1 � G12
1 Introduction
The controversy surrounding the abnormality of stock prices has been the subject of
extensive research over the past few decades. Essentially two competing mutually
T. Choudhry (&) � R. JayasekeraSchool of Management, University of Southampton, Southampton SO17 1BJ, UKe-mail: [email protected]
R. Jayasekerae-mail: [email protected]
123
Rev Quant Finan AccDOI 10.1007/s11156-013-0404-6
independent hypotheses have emerged, each explaining certain aspects of stock price
behaviour.
The first, which we describe as ‘asset mispricing’, puts forth a behavioural finance
argument to explain certain anomalies of stock price behaviour. These studies present an
explanation of the evident over/under reaction of stock prices to information This essen-
tially suggests market inefficiency. De Bondt and Thaler (1989), Chopra et al. (1992),
Ritter (1991), Loughran and Ritter (1995), Spiess and Affleck-Graves (1995), and Dharan
and Ikenberry (1995) all present evidence of market inefficiency in terms of overreaction to
information. Evidence of under-reaction is just as frequent, as shown by Ball and Brown
(1968), Bernard and Thomas (1990), Jegadeesh and Titman (1993), Cusatis et al. (1993),
Desai and Jain (1997), Ikenberry et al. (1996), Lakonishok and Vermaelen (1990), Iken-
berry et al. (1995), Michaely et al. (1995), Asquith (1983), Agrawal et al. (1992), Roll
(1986), Ikenberry and Lakonishok (1993) and, more recently, Frazzini (2006).
The alternative argument for ‘‘market efficiency’’ serves to enforce the efficient market
hypothesis; Fama (1970, 1991), Fama and French (1992, 1993, 1998, 2002).1 Chan (1988)
and Ball and Kothari (1989), all provide evidence that the beta of individual stock rises
(falls) in response to abnormally negative (positive) returns, and argue that this asymmetric
response to good and bad news explains the performance of stock returns, i.e. there exists
predictive asymmetry in conditional betas’ response to shocks. Ball and Kothari (1989)
argue that this asymmetric response to good and bad news explains the performance of
over/under reaction experienced by the ‘‘winners’’ and ‘‘losers’’ stocks. They show that in
an efficient market time-varying expected returns are caused by: a variation in expected
returns on the market portfolio; the relative risk of a firm’s investments; and leverage.
Thus, if the firm beta changes asymmetrically in response to news (shocks) this provides
support for the efficient market hypothesis (Cho and Engle 1999).2 Therefore the detection
of asymmetry in beta lends support to the market efficiency theory as the actual degree of
mispricing is now less because some of it can be explained by the change in beta.
Our research views this controversy from a different perspective, through the analysis of
stock returns of UK firms leading up to and during the financial crisis of 2007–2010.3 Thus,
we investigate the asymmetric effect of beta during good and bad periods to good and bad
news, and shed fresh insight to the controversy of the ‘‘abnormalities of stock prices’’, i.e.
whether the hypothesis of ‘increased market efficiency’ or ‘asset mispricing’ would better
fit the empirical observations as an economy slides from a boom to a rescission. We focus
on the UK equity market, analysing the relative effects on five major industrial segments.
Our main hypothesis in this paper is to investigate whether any asymmetric effect exist in
betas of UK firms during the pre-crisis and the crisis periods? The bivariate BEKK
GARCH model is employed to estimate the time-varying beta for each firm. The theory of
time-varying beta is based on Bodurtha and Mark (1991). Then linear regression is used to
capture the effects of a period when the UK economy slid from relative prosperity (pre-
crisis) to a recession, i.e. the current financial crisis period. We define the pre-crisis period
(which we refer to as the ‘‘good’’ period) from January 2004 to June 2007 and the crisis
period (which we refer to as the ‘‘bad’’ period) to commence from June 2007 to September
2010. Thus, we investigate 7 years of daily data, classified under 5 major industries
1 Harel et al. (2011) provide an analysis of efficient markets.2 Cho and Engle (1999), finding evidence of asymmetric effects in betas of US firms, claim that this impliesthat abnormities of stock prices can be explained by changes in expected returns through a change in beta,thus supporting the claims of Chan (1988) and Ball and Kothari (1989).3 Of course, the crisis has carried on beyond 2010.
T. Choudhry, R. Jayasekera
123
covering 25 companies in total. The industries under study are: Banking, Retail, Food,
Construction and Oil. Although there is an extensive body of literature on the controversy
surrounding asset mispricing and market efficiency, to the best of our knowledge there is
no published research4 which looks at this puzzle during this financial crisis period, and
thus our findings make a valuable contribution in this area. Also, given the lack of research
in this area for the UK this paper makes a vital contribution to the literature.
We proceed by describing and explaining briefly the global financial crisis of
2007–2010 in Sect. 2. The conditional CAPM and the time-varying beta are presented in
Sect. 3. In Sect. 4 we describe the data and the BEKK GARCH modelling framework
employed. The GARCH results are briefly described in Sect. 5. Section 6 explains the
theoretical underpinnings of time varying beta and the general framework that we employ
to capture the effect of good and bad news leading up to and during the financial crisis
period. In particular we explain how we interpret the asymmetric betas, justifying our
arguments. The asymmetric effect results and their interpretation are in Sect. 7. We con-
clude in Sect. 8.
2 Global financial crisis of 2007–2010
The current financial crisis that first hit the global economy in the summer of 2007 is
without precedent in post-war economic history. However, although its size and extent
are exceptional, the crisis still has many features in common with similar financial-stress
driven recession episodes in the past. The crisis was preceded by a long period of rapid
credit growth, low risk premiums, abundant availability of liquidity, strong leveraging,
soaring asset prices and the development of bubbles in the real estate sector. Over-
stretched leveraging positions rendered financial institutions extremely vulnerable to
corrections in asset markets.5 As a result, a turn-around in a relatively small corner of the
financial system (the US subprime market) was sufficient to topple the whole structure.6
Such episodes have happened before (e.g. Japan and the Nordic countries in the early
1990s, the Asian crisis in the late-1990s, the US S&L crisis in the mid-1980s), however
this time is different as the crisis has had a truly global affect (European Commission
2009).
Initially the UK companies affected were those directly involved in home construction
and mortgage lending such as Northern Rock7and Countrywide Financial, as they could no
longer obtain financing through the credit markets.
4 Veronesi (1999) presents a two state continuous time hidden Markov chain model to explain the stockmarket under-reaction to bad news in good times. We use a BEKK GARCH framework analysing the impactof good and bad news in good and bad periods and present empirical evidence for 5 major industryclassifications leading up to and during the recent financial crisis period.5 Dwyer and Lothian (2012) state that cross-country evidence and analyses of individual countries suggest acommon explanation to the cause of the financial crisis is likely to be based in rapid credit expansion andeconomic growth. Dias and Ramos (2013) study the behaviour of the banking sector of 40 countries duringthe period 2007–2010. They show that although there were periods of intense contagion, the impact wasuneven among sample countries. Marsh and Pfleiderer (2012) provide a discussion of black swans and thefinancial crisis.6 However, Kamin and DeMarco (2012) conclude that issues with U.S. Sub Prime mortgages more plau-sibly were a wake-up call about banking problems around the world than a direct cause of those problems.7 See Shin (2009). Reflections on Northern Rock.
Level of efficiency in the UK equity market
123
The above graph demonstrates the behaviour of the FTSE 100 index, which can be
regarded as a proxy for the UK economy during the period under consideration. As is
evident, there is a marked decline8 in the index along with its market capitalisation
commencing from the second half of 2007.9 By August 2007 the European Central Bank
injected 170bn euros into the banking market and the Fed had lowered interest rates in an
attempt to revive the credit markets. September 2007 saw the fall of Northern Rock
(Dwyer and Tkac 2009), a UK building society, as it received emergency financial support
from the Bank of England. Shares in Northern Rock fell by 32 % after it emerged that it
had approached the Bank of England for help. September 2008 witnessed Lloyds bank
make a 12.2 bn takeover of the ailing Halifax Bank of Scotland (HBOS), the UK’s largest
mortgage lender, after its shares plummeted amid concerns over the firm’s future. The UK
government invoked a national interest clause to bypass competition law, as this merger
was responsible for close to one-third of the UK’s savings and mortgage market. October
2008 saw the UK government injected £37 billion in an attempt to rescue RBS and Lloyds-
HBOS as the as financial markets collapsed. In April 2009 the UK Chancellor Alistair
Darling revealed that the credit crunch would lead to the largest budget deficit in UK
financial history of £175 bn, with total government debt set to double to £1 trillion by
2014.10 Mr Darling admitted that it will take approximately 10 years to get the budget back
to the position it was in before the credit crunch.11 These events are plotted along the
timeline by the yellow bars across the FTSE 100 index movement.
Turbulent financial market conditions can result in the correlation between the returns
of financial assets to ‘‘breakdown’’. Increasing correlation during volatile market condi-
tions implies a reduction in portfolio diversification benefits and will have obvious
Source: Bloomberg
8 In autumn 2008, financial markets did move very much in sync, with stock prices around the world fallingby 30 % or more (Bartram and Bodnar 2009).9 On average the Index declined by approximately 10 % between the periods 2004 to 2007 and 2007 to2010.10 These figures were sourced from the financial times and the BBC website.11 However, sentence et al. (2012) state that the substantial increase in the UK house prices and capitalinflows associated with growth of private sector debt combined with a large financial sector exposed toforeign developments led many observers to expect a worse experience than has transpired.
T. Choudhry, R. Jayasekera
123
implications in portfolio asset allocation. Table 1a presents the correlation between the five
industrial returns during the pre-crisis period and the crisis period. Similarly Table 1b
shows the covariance between the industrial returns during the pre-crisis period and the
crisis period.12 The top half of both tables show results from the pre-crisis period and the
bottom halves (the italicized values) present results from the crisis period. As stated earlier
industries under study are: Banking, Retail, Food, Construction and Oil. The tables were
constructed from the sector indices sourced from DataStream for the pre crisis and the
crisis period. The analysis of the correlation of returns provides a simple indicator of the
co-movements of stock indices or a rough measure of market interdependence. Thus, the
size and evolution of the correlation between equity markets is important for proper
diversification. The correlation coefficients are all greater than zero and varying in size
depending on the industries (Table 1a). The highest correlation is larger than 0.6 and the
lowest is less than 0.3. The results indicate that correlation between these five industries
increases to some extend from the pre-crisis to the crisis period. For example, the corre-
lation between Banking industry and oil industry increases from 0.447 to 0.531 and
between the industries of construction and Retail, the correlation increases to 0.659 from
0.575. Table 2b indicates positive covariance among all combinations of the returns during
both periods and this result implies that returns behaviour in a similar manner. The
covariance tends to fall from the pre-crisis to the crisis period. For example the covariance
between the industries of Banking and Construction falls from 0.0004 to 0.00032 and
between Retail and Oil to 0.0002 from 0.00031.
3 The (conditional) CAPM and time-varying beta
One of the assumptions of the capital asset pricing model (CAPM) is that all investors have
the same subjective expectations on the means, variances and co-variances of returns.13
12 We thank the referee for suggesting the correlation and covariance tests.13 See Markowitz (1952), Sharpe (1964) and Lintner (1965) for details of the CAPM.
Table 1 A correlation between industrial returns before and during the crisis period
Pre-crisis period 1 Jan 2004–2030 June 2007
Banks Construction Oil Food Retail
a
Banks 1.00 0.536 0.447 0.482 0.589
Construction 0.575 1.00 0.417 0.416 0.575
Oil 0.531 0.539 1.00 0.362 0.383
Food 0.422 0.490 0.546 1.00 0.465
Retail 0.643 0.659 0.473 0.500 1.00
b
Banks – 0.000040 0.000034 0.000027 0.000035
Construction 0.00032 0.000041 0.000030 0.000045
Oil 0.00032 0.00019 – 0.000027 0.000031
Food 0.00019 0.00013 0.00016 – 0.000027
Retail 0.00041 0.00025 0.00020 0.00027 –
Crisis period 1 July 2007–2030 Oct 2010 (italicized values)
Level of efficiency in the UK equity market
123
According to Bollerslev et al. (1988), economic agents may have common expectations on
the moments of future returns, but these are conditional expectations and therefore random
variables rather than constant.14 The CAPM that takes conditional expectations into con-
sideration is sometimes known as conditional CAPM. This conditional CAPM provides a
convenient way to incorporate the time-varying conditional variances and co-variances
(Bodurtha and Mark 1991).15 An asset’s beta in the conditional CAPM can be expressed as
the ratio of the conditional covariance between the forecast error in the asset’s return, and
the forecast error and the conditional variance of the forecast error in the market return.
14 According to Klemkosky and Martin (1975), betas will be time-varying if excess returns are charac-terized by conditional heteroscedasticity.15 Hansen and Richard (1987) have shown that omission of conditioning information, as is done in tests ofconstant beta versions of the CAPM, can lead to erroneous conclusions regarding the conditional meanvariance efficiency of a portfolio.
Table 2 Firm description
Company Abbreviation Industry Total sharesoutstanding(millions)
Market capitalization(£ million) asat 10/01/2011
Kingfisher KGF Retail 2361.974 6,190.73
Marks and Spencer MKS Retail 1583.644 6,092.28
Next NXT Retail 183.321 3,792.92
Inchcape INCH Retail 460.505 1,780.31
Home Retail group HOME Retail 818.633 1,665.1
Barclays BARC Banking 12,181.94 33,591.7
HSBC HSBA Banking 17,686.16 118,921.7
Lloyds LLOY Banking 68,074.13 44,328.05
Royal Bank of Scotland RBS Banking 58,458.13 43,225.02
Standard Charter STAN Banking 2,348.155 40,165.19
British Petroleum BP Oil 18,796.54 91,285.4
Royal Dutch RDSB Oil 3,565.953 133,812
BG Group BG Oil 3,386.378 44,886.44
Tullow TLW Oil 888.237 11,991.2
ENSCO ESV Oil 143.397 4,707.806
James Halstead JHD Land and construction 51,941 389.56
SEGRO SGRO Land and construction 741.537 2,126.73
Keller KLR Land and construction 64.311 420.91
Kier Group KIE Land and construction 37.906 514.77
Capital ShoppingCentres Group
CSCG Land and construction 692.673 2,718.05
Tesco TSCO Food and retail 8,029.803 34,500.05
Sainsbury SBRY Food and retail 1,865.939 7,277.16
Morrison MORW Food and retail 2,657.746 7,221.1
Associated British Foods ABF Food and retail 791.674 9,040.92
Unilever ULVR Food and retail 1,310.156 57,760.29
T. Choudhry, R. Jayasekera
123
The following analysis relies heavily on Bodurtha and Mark (1991) and Choudhry and
Jayasekera (2012). Let Ri,t be the nominal return on asset i (i = 1, 2,…,n) and Rm,t the
nominal return on the market portfolio m. The excess (real) return of asset i and the market
portfolio over the risk-free asset return is presented by ri,t and rm,t respectively. The
conditional CAPM in excess returns may be given as
E ri;tjIt�1
� �¼ biIt�1E rm;tjIt�1
� �ð1Þ
where,
biIt�1 = cov Ri;t; Rm;tjIt�1
� �=var Rm;tjIt�1
� �¼ cov ri;t; rm;tjIt�1
� �=var rm;tjIt�1
� �ð2Þ
and E(|It-1) is the mathematical expectation conditional on the information set available to
the economic agent’s last period (t-1), It-1. Expectations are rational, based on Muth’s
(1961) definition of rational expectation, where the mathematical expected values are
interpreted as the agent’s subjective expectations. According to Bodurtha and Mark (1991),
asset i risk premium varies over time due to three time-varying factors: the market’s
conditional variance; the conditional covariance between the asset’s return; and the mar-
ket’s return and/or the market’s risk premium.
The asymmetric effect16 of news on the volatility of stock returns has been investigated
and evidenced in many past studies (Black 1976; French et al. 1987; Nelson 1991; Schwert
1989). The effect refers to the volatility trends in individual stocks and market indices
where one can observe a rise in volatility following negative returns and a fall following
positive returns. The effect can be rationalised in terms of a leverage (financial and
operational) based explanation, or one based on the determinants of market risk premium.
The former stems from the notion of viewing equity as a call option on the value of the
firm’s assets where the option becomes worthless when the asset value falls below the
liabilities (i.e. the strike price). Thus, if the value of a leveraged firm drops, its equity
becomes highly leveraged, causing an increase in volatility17 (Black 1976; Christie 1982).
The second explanation stems from the positive relationship between volatility (which is a
proxy for risk) and the expected market risk premium (the expected return on a stock
portfolio minus the riskless rates). Under the assumption of a rational investor paradigm,
ceteris parabus, an increase in volatility increases the expected return which in turn lowers
the stock price contributing to the asymmetric effect in volatility (Pindyck 1984; Poterba
and Summers 1986; French et al. 1987; Bollerslev et al. 1988; Engle et al. 1990; Campbell
and Hentschel 1992).
Asymmetry in volatility may also imply asymmetry in time-varying beta. If the risk
premium is an increasing function of the volatility, and the beta is a proxy for risk, then the
asymmetric effect in volatility may imply such an effect occurs in beta too18 (Cho and
Engle 1999). Furthermore, if the beta of a leveraged firm’s asset is positive, the beta of the
firm’s equity should rise in response to negative returns, as the firm takes on more leverage.
Thus, the expected equity betas tend to be increasing with leverage. Further, Braun et al.
16 This is also referred to as the ‘‘leverage effect’’.17 Christie (1982) shows that equity volatility is increasing in financial leverage, and hence there is anegative relationship between the variance of returns and the value of equity. However, Christie (1982) andBlack (1976) point out that financial and operational leverage is not enough to fully account for theasymmetry of volatility.18 According to Brooks and Henry (2002), if the risk premium is increasing in volatility, and if beta is aproper measure of the sensitivity to risk, then time variation and asymmetry in the variance–covariancestructure of returns may lead to time variation and asymmetry in beta.
Level of efficiency in the UK equity market
123
(1995) and Ball and Kothari (1989), claim that an increase (decrease) in market shocks to
the firms also increases (decreases) beta and leads to a rise (fall) in expected returns on the
market. This should result in a drop in the stock price contributing to the asymmetric
effect. In this paper we investigate whether any asymmetric effect exists in betas of UK
firms during the pre-crisis and crisis periods.
4 BEKK GARCH modelling framework and the data
As stated earlier, the time-varying beta of each firm in this paper is estimated by means of
the BEKK GARCH model.19 The following bivariate GARCH(p,q) model may be used to
represent the returns from asset i and the market portfolio (m). This presentation is termed
by Engle and Kroner (1995) the BEKK model; the conditional covariance matrix is
parameterized as20
yt ¼ lþ et � het�1 ð3Þ
et=Xt�1�N 0; Htð Þ ð4Þ
vech Htð Þ ¼ C0C þXK
K¼1
Xq
i¼1
A0Kiet�ie0t�iAki þ
XK
K¼1
Xp
i¼1
B0KjHt�jBkj ð5Þ
where yt = (rit, rm,t) is a (2 9 1) vector containing excess returns from asset i and the
market portfolio (m), l is a 2 9 1 vector of constant, Aki, i = 1,…,q, k = 1,…K, and Bkj
j = 1,…p, k = 1,…,K are all N 9 N matrices. This formulation has an advantage over the
general specification of the multivariate GARCH in that conditional variance (Ht) is
guaranteed to be positive for all t (Bollerslev et al. 1992). The moving average (MA) term
het-1 is included to capture the effect of non-synchronous trading. According to Susmel
and Engle (1994), non-synchronous trading induces negative serial correlation, and the MA
term allows for autocorrelation induced by discontinuous trading in the asset.
The time-varying beta (b) for asset i is calculated as
bi;t ¼ H12;t=H22;t: ð6Þ
where H12;t is the estimated conditional covariance between the specific asset returns and
market portfolio returns, and H22;t is the estimated conditional variance of the market
portfolio returns from the bivariate BEKK GARCH model. Given that conditional
covariance and conditional variance are time-dependent, the stock beta will be time-
dependent. The time-varying beta defined in Eq. 6 is applied in this paper.21
Daily stock price indices from twenty five individual firms from the UK are applied in
the tests. The data range from January 1, 2004 to October 30, 2010. The total period is
further broken into the pre-crisis period (1 Jan 2004–30 June 2007) and the crisis period (1
July 2007–30 October, 2010). Table 2 presents the basic information about the firms, the
size of the firm and the industry they belong in. Firms under study are chosen based on the
criteria of size and industry. The size of the firm is based on market capitalisation. The
large variation in the size of the firms and the industries they belong to is clearly visible.
19 Thus we estimate the BEKK model for each firm to create 25 individual time varying betas.20 The BEKK description relies heavily on Choudhry and Jayasekera (2012).
21 We estimate the BEKK GARCH to obtain H12;t and H22;t for each firm to estimate the betas.
T. Choudhry, R. Jayasekera
123
The FTSE All stock index was used as a proxy for the market portfolio. Stock returns are
simply created by taking the first difference of the log of the stock index. The return on the
risk-free asset is represented by the return on the three-month UK Treasury bill. The excess
stock return is calculated as the nominal stock returns minus the returns on the bill. All data
are taken from Datastream International.
5 The bivariate BEKK GARCH results
Given the bulkiness of the BEKK results they are not provided to conserve space but they
are available on request. We provide a basic summary of the BEKK results. The BEKK
bivariate GARCH results are quite standard. The ARCH coefficients (A11 and A22) are all
positive and significant implying volatility clustering in both the firm and the market
returns. All ARCH coefficients are less than unity in size. In all models the GARCH
coefficient is significant and positive implying the GARCH effect. A large coefficient of
the GARCH term indicates that shocks to conditional variance take a long time to die out
and volatility persists. Not much evidence is found indicating a linkage between the
volatilities (A12 and A21) and conditional variances (B12 and B21) of the firm and the
market. Some evidence of non-synchronous trading (h) is found, mostly in the firms. The
significant h are mostly positive. To assess the general descriptive validity of the model, a
battery of standard specification tests is employed. Specification adequacy of the first two
conditional moments is verified through the serial correlation test of white noise. These
tests employ the Ljung-Box Q statistics on the standardised (normalised) residuals
(et=H1=2t ) and standardised squared residuals (et=H2
t ). All series are found to be free of
serial correlation (at the 5 % level). The absence of serial correlation in the standardised
squared residuals implies the lack of need to encompass a higher order ARCH process
(Giannopoulos 1995).
Figure 1 presents five of the estimated betas of the twenty five UK firms. The shaded
region shows the crisis period (2007–2010). Change of movement of the beta before the
crisis and during the crisis is clearly visible, especially in the cases of the banking and retail
industries. Besides the basic movement not much can be deduced and analyzed from the
graphs. Graphs of other firm’s betas provide a somewhat similar story and are available on
request.
Table 3 shows the basic statistics of the betas during the pre-crisis (2004–2007) and
crisis period (2007–2010). The mean of most betas are found to be more than unity,
implying that most of the firms under study are more risky than the market. Only in the
food industry does each firm show less riskiness than the market during both periods. For
most firms the beta increases from the pre-crisis to the crisis period. This is as expected,
and this jump in the betas is especially obvious in the banking, retail and food industries. In
addition, most betas are found to be significantly skewed and/or leptokurtic and thus are
found to be non-normal by means of the Jarque–Bera statistics. This result is not unique, as
Choudhry (2002) also provides evidence of non-normal UK firm daily betas. Application
of the OLS requires that all variables are stationary. The stochastic structure of all twenty
five betas is investigated by means of the augmented Dickey-Fuller root test (ADF).22 All
beta are found to be stationary in levels during both periods. This result is not unique
22 The ADF tests are applied with six lags maximum.
Level of efficiency in the UK equity market
123
either, as Choudhry (2002) provides similar results for other UK firms, and Brooks and
Henry (2002) for the UK industrial sectors.
6 The rational and the general approach
Given the evidence on the predictive asymmetry of volatility, we investigate the asym-
metric effect of beta during good and bad periods to good and bad news in the UK equity
market during the pre-crisis and crisis period in order to shed fresh insight on the con-
troversy of the ‘‘abnormalities of stock prices’’. i.e. whether the hypothesis of ‘increased
market efficiency’ or ‘asset mispricing’ would better fit the empirical observations as an
economy slides from a boom to a rescission.
In particular we investigate the following hypothesis. Does any asymmetric effect exist
in betas of UK firms during the pre-crisis and crisis periods? Absence or presence of the
asymmetric effect in betas can lead to two different conclusions. The absence of any
asymmetric effect leads one to conclude that abnormalities of stock return (instances of
asset mispricing) are evidenced, thus demonstrating symptoms of market inefficiency.23
The presence of the asymmetric effect leads one to conclude that abnormalities of stock
returns (instances of asset mispricing) can, at least partially, be explained by changes in
expected returns through a change in beta. If the markets are efficient one would expect to
witness an asymmetric time varying beta in response to good and bad news. This is due to
the following reasons. Firstly, good or bad news both have an asymmetric effect on the
volatility of the stock prices. The asymmetric effect of news on the volatility of stock
returns is well documented (Black 1976; French et al. 1987; Nelson 1991; Schwert 1989).
One can observe a rise/fall in volatility following negative/positive news.24 Secondly, risk
premium is an increasing function of volatility, and the beta is a proxy for risk, thus the
asymmetric effect in volatility may imply such an effect occurs in beta too (Cho and Engle
1999 and Brooks and Henry 2002). If the beta of a leveraged firm’s asset is positive, the
beta of the firm’s equity should rise in response to negative returns, as the firm takes on
more leverage. Thus, the expected equity betas tend to be increasing with leverage just as
does the volatility. Braun et al. (1995) and Ball and Kothari (1989), claim that an increase
(decrease) in market shocks to the firms also increases (decreases) beta and leads to a rise
(fall) in expected returns on the market. This results in a drop in the stock price contrib-
uting to the asymmetric effect. Thus, the presence of market efficiency implies an
asymmetry in the time-varying beta. Conversely an absence of asymmetry indicates market
inefficiency via asset mispricing.
In this paper, the time-varying beta of individual UK firms’ stock returns is investigated
in the context of an asymmetric effect of news, market shocks and idiosyncratic shocks.
23 Previous research by Braun et al. (1995) support the overreaction theory (asset mispricing) by finding aweak asymmetric effect in beta. They conclude, based on the evidence of the low frequency (weekly) data,that betas are not responsive enough to account for the differing return performances of ‘‘winners’’ and‘‘losers’’, and thus support De Bondt and Thaler (1989).24 There are two possible explanations. (1) Leverage based—viewing equity as a call option for the firm’sassets—if the value of a leveraged firm drops, its equity becomes highly leveraged, causing an increase involatility (Black 1976; Christie 1982). (2) Positive relation between volatility and the expected market riskpremium—an increase in volatility increases the expected return which in turn lowers the stock pricecontributing to the asymmetric effect in volatility (Pindyck 1984; Poterba and Summers 1986; French et al.1987; Bollerslev, Engle and Wooldridge, 1988; Engle, Ng and Rothschild, 1990; Campbell and Hentschel1992).
T. Choudhry, R. Jayasekera
123
Barclays - Banking Industry
2004 2005 2006 2007 2008 2009 2010-1.0-0.50.00.51.01.52.02.53.03.5
Kingfisher - Retail Industry
2004 2005 2006 2007 2008 2009 2010-1.0-0.50.00.51.01.52.02.53.03.5
BP - Oil Industry
2004 2005 2006 2007 2008 2009 2010-1.0-0.50.00.51.01.52.02.53.03.5
Sainsbury - Food Industry
2004 2005 2006 2007 2008 2009 2010-1.0-0.50.00.51.0
1.52.02.53.03.5
Keller - Construction Industry
2004 2005 2006 2007 2008 2009 2010-1.0-0.50.00.51.01.52.02.53.03.5
Fig. 1 Time varying betas
Level of efficiency in the UK equity market
123
Ta
ble
3B
eta
bas
icst
atis
tics
Fir
ms
Pre
-cri
sis
20
04–
20
07
Cri
sis
per
iod
20
07
–2
01
0
Mea
nV
aria
nce
Sk
ewnes
sK
urt
osi
sJ–
BA
DF
Mea
nV
aria
nce
Sk
ewnes
sK
urt
osi
sJ–
BA
DF
Ban
ks
Bar
clay
s1.1
65
a0
.05
60
.17
5a
0.8
31
a3
0.9
4a
-5
.945
**
*1
.54
0a
0.1
25
0.8
91
a0
.805
a1
35
.40
a-
4.4
54
**
*
HS
BC
0.9
17
a0
.02
90
.35
9a
0.6
67
a3
6.5
9a
-5
.038
**
*1
.02
6a
0.0
33
-0
.042
0.0
33
0.2
90
-4
.51
9*
**
Llo
yd
s1
.050
a0
.09
81
.93
3a
7.0
66
a2
,468
.8a
-5
.387
**
*1
.33
2a
0.1
78
0.0
60
1.4
34
a7
3.2
8a
-4
.97
8*
**
Sta
nd
ard
1.1
92
a0
.07
20
.42
5a
0.5
46
a3
8.8
9a
-4
.474
**
*1
.32
1a
0.0
65
1.1
06
a1
.508
a2
53
.58
a-
4.1
94
**
*
RB
S1
.112
a0
.06
30
.71
7a
7.4
90
a2
,212
.2a
-7
.929
**
*1
.54
7a
0.2
51
1.9
60
a5
.920
a1
,78
3.6
a-
7.0
27
**
*
Ret
ail
Kin
gfi
sher
1.0
00
a0
.05
40
.47
1a
1.4
66
a1
15
.51
a-
4.8
83
**
*1
.15
4a
0.1
12
1.0
06
a0
.919
a1
73
.00
a-
3.3
33
**
*
M&
S0
.826
a0
.07
6-
0.2
08
b2
.653
a2
74
.28
a-
12
.80
8*
**
0.9
20
a0
.13
4-
0.2
13
b5
.664
a1
,14
1.6
a-
13
.19
2*
**
Nex
t0
.805
a0
.04
6-
0.7
68
a1
.777
a2
09
.82
a-
5.0
02
**
*1
.10
0a
0.0
46
1.1
67
a1
.051
a2
31
.63
a-
2.7
59
**
*
Inch
0.8
54
a0
.06
9-
0.0
72
0.6
19
a1
5.1
3a
-1
1.7
47
**
*0
.96
0a
0.1
34
-0
.135
0.2
86
5.4
47
-1
2.4
79
**
*
Ho
mer
etai
l0
.865
a0
.05
20
.08
50
.182
2.3
67
-4
.250
**
*1
.14
7a
0.0
74
1.2
22
a1
.001
a2
47
.22
a-
3.2
80
**
*
Oil B
P1
.054
a0
.03
80
.15
0-
0.0
84
3.7
10
-3
.903
**
*0
.91
7a
0.0
24
-0
.588
a0
.731
a6
7.9
0a
-3
.30
0*
**
Roy
al1
.104
a0
.05
41
.27
2a
4.3
00
a9
48
.31
a-
5.9
74
**
*0
.95
4a
0.0
24
-1
.131
a1
.436
a2
54
.07
a-
3.4
82
**
*
BG
0.9
19
a0
.03
40
.54
0a
6.5
22
a1
,662
.5a
-9
.310
**
*1
.04
0a
0.0
42
1.3
06
a8
.306
a2
,67
5.7
a-
9.3
01
**
*
To
llow
1.1
82
a0
.18
6-
0.0
26
-0
.460
a8
.170
a-
3.7
16
**
*1
.02
8a
0.0
67
-0
.507
a2
.153
a2
00
.27
a-
4.6
53
**
*
EN
SC
O0
.670
a0
.22
8-
0.0
23
0.2
07
1.7
32
-3
.617
**
*0
.80
0a
0.0
89
-0
.416
a-
0.4
03
b3
0.2
99
a-
3.0
04
**
Fo
od
TE
SC
O0
.590
a0
.03
7-
0.2
86
a0
.605
a2
6.4
4a
-6
.737
**
*0
.80
3a
0.0
31
-0
.870
a0
.926
a1
37
.63
a-
8.3
60
**
*
Sai
nsb
ury
0.6
15
a0
.07
8-
1.3
25
a8
.324
a2
,914
.6a
-6
.900
**
*0
.88
1a
0.1
06
-7
.930
a1
00
.49
a3
,66
0.8
a-
8.3
79
**
*
Mo
rris
on
0.5
42
b0
.10
4-
3.5
80
a3
3.9
9a
45
,89
7.0
a-
6.6
50
**
*0
.85
4a
0.0
25
1.8
75
a1
0.5
82
a4
,45
9.2
a-
7.0
60
**
*
Ass
oB
riti
sh0
.319
a0
.05
4-
0.0
45
-0
.084
0.5
85
-3
.490
**
*0
.78
3a
0.0
18
-1
.967
a1
1.3
26
a5
,08
5.9
a-
3.4
90
**
*
Un
ilev
er0
.748
a0
.04
5-
0.0
76
0.0
40
0.9
50
-5
.960
**
*0
.78
4a
0.0
44
-1
.713
a3
.750
a9
14
.92
a-
5.2
10
**
*
T. Choudhry, R. Jayasekera
123
Ta
ble
3co
nti
nued
Fir
ms
Pre
-cri
sis
20
04–
20
07
Cri
sis
per
iod
20
07
–2
01
0
Mea
nV
aria
nce
Sk
ewnes
sK
urt
osi
sJ–
BA
DF
Mea
nV
aria
nce
Sk
ewnes
sK
urt
osi
sJ–
BA
DF
Co
nst
ruct
ion
Kel
ler
0.4
98
b0
.17
31
.02
94
.067
a7
90
.52
a-
6.0
56
**
*0
.84
6a
0.0
60
0.3
30
a2
.600
a2
53
.96
a-
6.1
35
**
*
Kie
r0
.915
a0
.12
60
.94
4a
1.4
40
a2
14
.48
a-
3.6
33
**
*0
.74
8a
0.0
54
0.3
68
a-
0.5
59
a3
0.2
70
a-
3.2
61
**
Hal
stea
d0
.278
a0
.08
7-
0.4
94
a0
.410
b4
3.5
00
a-
3.8
32
**
*0
.47
0a
0.0
60
0.2
67
a-
0.9
49
a4
1.9
9a
-2
.68
8*
Seg
ro0
.841
a0
.10
6-
1.7
30
a1
0.9
60
a5
,024
.5a
-5
.500
**
*1
.10
6a
0.0
69
-0
.248
a-
0.3
01
11
.92
2a
-4
.21
0*
**
Cap
ital
1.0
00
a0
.08
30
.12
8-
0.8
61
a3
0.7
01
a-
3.4
11
**
1.1
33
a0
.06
00
.482
0.5
86
a4
5.0
21
a-
4.3
69
**
*
JBJa
rqu
e–B
era,
AD
FA
ug
men
ted
Dic
key
-Fu
ller
test
a,b
and
cS
ignifi
cantl
ydif
fere
nt
from
zero
atth
e1,
5an
d10
%le
vel
,re
spec
tivel
y
***,
**
and
*R
ejec
tion
of
the
null
of
unit
root
atth
e1,
5an
d10
%le
vel
resp
ecti
vel
y
Level of efficiency in the UK equity market
123
Focusing on these two shocks, we apply a model that allows one to distinguish two shocks
in the beta process. The double beta model specification is used for parsimonious esti-
mation and computability. In this model, market information is used as an explanatory in
the estimation of the volatility and beta of the individual stock returns.
We apply the joint model based on Cho and Engle (1999) and Choudhry and Jayasekera
(2012) detailed below to each individual firm for the bad and good periods as follows. We
use the joint model for conditional time-varying beta (bi,t) which captures both the sys-
temic (market) effects as well as the non market, idiosyncratic effects, and is based on the
assumption that the beta follows an AR (1) process.25
For the good period
bi;tðG:PÞ ¼ cbðG:PÞ þ abðG:PÞ bi;t�1ðG:PÞ � cbðG:PÞ
� �þ diðG:PÞzi;t�1ðG:PÞ
þ dmðG:PÞzm G:Pð Þ;t�1ðG:PÞþet ð7Þ
For the bad period
bi;tðB:PÞ ¼ cbðB:PÞ þ abðB:PÞ bi;t�1ðB:PÞ � cbðB:PÞ
� �þ diðB:PÞzi;t�1ðB:PÞ
þ dmðB:PÞzm B:Pð Þ;t�1ðB:PÞþet ð8Þ
where zi = non-market Shocks, zm = market Shocks, di = Coefficient of the non market
shocks. It shows the level of contribution from the non-market shocks towards the firms
individual beta, dm = Coefficient of the market shocks. It shows the level of contribution
from the market shocks towards the firms individual beta, et = standard error term with
zero mean and constant variance.
Following Cho and Engle (1999), the terms dizi and dmzm, allow for leverage (asym-
metric) effects in the time-varying betas based on non-market and market shocks
respectively. If di is negative and significantly different from zero, the beta (bi) will rise in
response to negative non-market returns (idiosyncratic returns), and fall in response to
positive non-market returns. Thus if di is significant and negative, it could be that there
exists a leverage effect via non-market shocks in the beta process. Similarly, if dm is
negative and significant, the beta rises in response to negative market returns, and falls in
response to positive market returns. In other words, if there is bad news in the market and
such shocks have an asymmetric effect; dm should be significant and negative.
The next step is to establish whether the systemic, market shocks or idiosyncratic, non
market shocks are prevalent in each individual firm using the log likelihood test for each of
the respective periods. The joint model of Eqs. 3 and 4 considers both the non-market and
the market shocks for the time-varying beta. Thus the log likelihood test is used to
determine whether the market shocks or the non market shocks or both are prevalent in the
joint model. Following this filtering process, when only the market shocks are prevalent we
obtain the ‘‘Market model’’, where only the non market shocks are prevalent we get the
‘‘non-market model’’, and the ‘‘joint model’’ is prevalent when both these shocks are
significant. Thus in order to arrive at the non-market model, markets shocks zm,t-1 are
omitted from the joint model by testing the significance of market shocks for the respective
periods.
For the good period for non market effects we test the null (H0) against the alternative
(H1) hypothesis as follows.
25 A zero order for AR in beta gives the beta extreme volatility implying complete stochastic behaviouranalogous to a random walk. Given that beta is a time-varying process, zero order for AR does not seem tobe a realistic model.
T. Choudhry, R. Jayasekera
123
H0 : dm; G:Pð Þ ¼ 0
H1 : dm; G:Pð Þ 6¼ 0
If the null hypothesis (H0) cannot be rejected, the non-market model is chosen, and it
implies that the beta process is driven only by non-market (idiosyncratic) shocks. Thus we
have the non-market model.
bi;tðG:PÞ ¼ cbðG:PÞ þ abðG:PÞ bi;t�1ðG:PÞ � cbðG:PÞ
� �þ diðG:PÞzi;t�1ðG:PÞþet ð9Þ
Similarly, the market model is obtained by estimating the significance of non-market
shocks zi,t-1 in the joint model. In this case the non-market shocks zi,t-1 are omitted from
the joint model and we arrive at the market model. The model selection between the joint
model and the market model is based on testing how the non-market volatility affects beta
in the joint model. For market effects on the overall firm beta we test the null (H0) against
the alternative (H1) hypothesis as follows.
H0 : di; G:Pð Þ ¼ 0
H1 : di; G:Pð Þ 6¼ 0
If the null hypothesis (H0) cannot be rejected, the market model is chosen, and it implies
that the beta is only driven by the market shocks. Thus we have the market model.
bi;tðG:PÞ ¼ cbðG:PÞ þ abðG:PÞ bi;t�1ðG:PÞ � cbðG:PÞ
� �þ dmðG:PÞzm G:Pð Þ;t�1ðG:PÞþet ð10Þ
we follow similar lines of reasoning for the bad period in arriving at the model selection.
Table 4 provides a summary of the model selections.
7 Asymmetric effects test results
Tables 5, 6, 7, 8, 9 presents the asymmetric effects test (models 1–4) results from the five
industries. For each firm two sets of results are presented, one for the pre-crisis period and
one for the crisis period. Each table presents the results from the joint model, the idio-
syncratic model and the market model. These models are estimated by means of ordinary
least squares, and then the covariance matrix estimates are corrected to allow for more
Table 4 Summary of the model selection
Market model (M.M) Non market model (N.M.M) Joint model (J.M)
Good period Test forH0: d i,, (G.P) = 0H1: d i, (G.P) = 0Select M.M if H0
cannot be rejected
Test forH0: dm,(G.P) = 0H1: dm,(G.P) = 0Select N.M.M if H0
cannot be rejected
Is applicable when H1 for boththe hypotheses for the goodperiod cannot be rejected
Bad period Test forH0: d i,, (B.P) = 0H1: d i, (B.P) = 0Select M.M if H0
cannot be rejected
Test forH0: dm,(B.P) = 0H1: dm,(B.P) = 0Select N.M.M if H0
cannot be rejected
Is applicable when H1 for boththe hypotheses for the badperiod cannot be rejected
Level of efficiency in the UK equity market
123
Table 5 Estimation of the Cho and Engle models (industry = banking)
Company a ab di dm H0: di = 0 H0: dm = 0
Barclays—pre-crisis period
Joint model 1.324a
(99.527)-0.0000(-0.3146)
-0.0592c
(-1.9269)0.0966b
(2.1781)– –
Idiosyncratic model 1.318a
(96.309)-0.0000(-0.3122)
-0.0172(-0.9032)
– 0.8158 –
Market model 1.335a
(101.704)-0.0000(-0.2742)
– 0.0339(1.5660)
– 2.4520
Barclays—crisis period
Joint model 1.493a
(81.962)-0.0000(-1.255)
-0.0203a
(-3.4563)0.0307(1.6021)
– –
Idiosyncratic model 1.484a
(86.055)-0.0000(-1.282)
-0.0133b
(-2.500)– 6.2270** –
Market model 1.507a
(84.890)-0.0000(-1.204)
– 0.0016(0.1033)
– 0.0107
HSBC—pre-crisis period
Joint model 0.852a
(143.120)0.0000(0.7890)
-0.0992a
(-5.4800)0.0589a
(3.4010)– –
Idiosyncratic model 0.848a
(139.130)0.0000(0.7005)
-0.0611a
(-5.8428)– 34.1380*** –
Market model 0.863a
(147.000)0.0000(0.5338)
– -0.0216b
(-2.066)– 4.2690**
HSBC—crisis period
Joint model 0.998a
(159.705)-0.0000(-1.088)
-0.0171a
(-4.7455)0.0027(0.4987)
– –
Idiosyncratic model 0.998a
(162.453)-0.0000(-1.119)
-0.0157a
(-7.0794)– 50.1180*** –
Market model 1.002a
(162.587)-0.0000(-1.3115)
– -0.0148a
(-4.0164)– 16.1319***
LLOYDS—pre-crisis period
Joint model 1.068a
(122.726)0.0000(0.670)
-0.0866a
(-4.3960)-0.0152(-0.7615)
– –
Idiosyncratic model 1.069a
(122.125)0.0000(0.733)
-0.0935a
(-6.6802)– 44.6250*** –
Market model 1.083a
(135.820)0.0000(0.3809)
– -0.0890a
(-6.1870)– 38.2776***
LLOYDS—crisis period
Joint model 1.156a
(108.197)0.0000(0.658)
-0.0126a
(-3.640)-0.0163c
(-1.7190)– –
Idiosyncratic model 1.163a
(117.426)0.0000(0.694)
-0.0151a
(-4.912)– 24.1320*** –
Market model 1.167a
(111.453)0.0000(0.572)
– -0.0330a
(-3.691)– 13.6260***
Royal bank of Scotland—pre-crisis period
Joint model 1.086a
(99.302)-0.0000(-1.622)
0.0268(0.6887)
-0.1336a
(-3.2129)– –
Idiosyncratic model 1.099a
(86.240)-0.0000(-1.616)
-0.0368(-1.2504)
– 1.5634 –
T. Choudhry, R. Jayasekera
123
robust (complex) behaviour of the residuals. In other words, the residuals are corrected for
autocorrelations.
Table 5 presents the results from the banking industry. In general the crisis period
seems to provide support for the market efficiency albeit to a lesser level compared to the
pre-crisis period. Based on the size of the coefficient (in absolute value) the level of
efficiency seems to decline from the pre-crisis to the crisis period. In the cases of HSBC
and Lloyds both periods require the application of the joint model. Thus the betas are
influenced by both the market and the non-market shocks. These two banks indicate
evidence of market efficiency during both periods. In the case of both banks the significant
coefficient on the non-market shock (di) is negative, implying an asymmetric effect due to
non-market shocks. Thus, for these firms the beta (bi) will rise in response to negative non-
market returns (idiosyncratic returns), and fall in response to positive non-market returns.
In absolute value, the size of the coefficient on the non-market shocks is much less than
unity, implying a small size effect of the non-market shocks on the beta. The coefficient on
the market shock provides a similar conclusion. For the remaining banks the model
selection indicates application of different models during the two periods. Application of
different models between the two periods implies to some extent the effect of the crisis. For
example, for the Royal Bank of Scotland during the pre-crisis period the market model is
Table 5 continued
Company a ab di dm H0: di = 0 H0: dm = 0
Market model 1.082a
(102.891)-0.0000(-1.626)
– -0.1099a
(-5.648)– 31.9010***
Royal bank of Scotland—crisis period
Joint model 1.380a
(79.395)0.0000(0.738)
0.0264a
(4.3378)-0.0663a
(-4.3000)– –
Idiosyncratic model 1.413a
(70.138)0.0000(0.372)
0.0211b
(2.5530)– 6.5174** –
Market model 1.354a
(79.194)0.0000(0.415)
– -0.0303a
(-2.7138)– 7.3650***
Standard charter—precrisis period
Joint model 1.185a
(116.546)-0.0000(-0.495)
-0.0618a
(-4.0320)0.0417c
(1.9552)– –
Idiosyncratic model 1.183a
(166.505)-0.0000(-0.507)
-0.043a
(-4.579)– 20.9630*** –
Market model 1.198a
(122.127)-0.0000(-0.568)
- -0.0273b
(-2.0923)– 4.3780**
Standard charter—crisis period
Joint model 1.273a
(115.612)-0.0000(-0.802)
-0.0251a
(-3.7140)0.0226b
(2.2644)– –
Idiosyncratic model 1.269a
(117.009)-0.0000(-0.845)
-0.0162a
(-3.5268)– 12.4381*** –
Market model 1.283a
(118.273)-0.0000(-0.905)
– -0.0104(-1.4316)
– 2.0494
t-statistics in the parenthesesa ,b and c Significantly different from zero at the 1, 5 and 10 % level, respectively
***, ** and * Rejection of the null hypothesis at the 1, 5 and 10 % level, respectively
Level of efficiency in the UK equity market
123
Table 6 Estimation of the Cho and Engle models (industry = retail)
Company a ab di dm H0: di = 0 H0: dm = 0
Kingfisher—pre-crisis period
Joint model 1.033a
(93.848)-0.0000(-0.1129)
-0.0244(-1.4208)
0.0409c
(1.9467)– –
Idiosyncratic model 1.030a
(92.881)-0.0000(-0.1411)
-0.0095(-0.7630)
– 0.5821 –
Market model 1.039a
(98.982)-0.0000(-0.1818)
– 0.0171(1.2653)
– 1.6000
Kingfisher—crisis period
Joint model 1.095a
(112.467)-0.0000(-0.9245)
-0.0394a
(-5.5346)0.0464a
(6.4505)– –
Idiosyncratic model 1.087a
(109.837)-0.0000(-1.0143)
-0.0191a
(-3.2893)– 10.8190*** –
Market model 1.114a
(121.276)-0.0000(-0.8131)
– 0.0109b
(1.9664)– 3.8666**
Marks and spencer—pre-crisis period
Joint model 0.5849a
(33.066)0.0000(1.3537)
-0.0663b
(-1.9648)-0.0070(-0.1533)
– –
Idiosyncratic model 0.5859a
(31.5937)0.0000(1.3586)
-0.0685a
(-2.6281)– 6.9066*** –
Market model 0.603a
(37.0035)0.0000(1.2045)
– -0.0517(-1.4388)
– 2.0702
Marks and spencer—crisis period
Joint model 1.0381a
(58.6119)-0.0000b
(-2.2723)0.0869b
(2.2580)-0.0816b
(-2.5357)– –
Idiosyncratic model 1.0649a
(34.095)-0.0000b
(-2.1491)0.0605c
(1.7180)– 2.9510* –
Market model 0.9975a
(86.0985)-0.0000b
(-2.1882)– -0.0087
(-1.1069)– 1.2250
Next—pre-crisis period
Joint model 0.8743a
(95.140)0.0000(0.0109)
0.0056(0.3729)
-0.0589a
(-2.8644)– –
Idiosyncratic model 0.8815a
(95.103)0.0000(0.2370)
-0.0145(-1.1490)
– 1.3206 –
Market model 0.873a
(98.387)0.0000(0.013)
– -0.0500a
(-3.5766)– 12.7918***
Next—crisis period
Joint model 1.1108a
(132.075)-0.0000(-0.1746)
-0.0287a
(-4.2634)0.0324a
(4.4611)– –
Idiosyncratic model 1.1048a
(131.159)0.0000(0.021)
-0.0140a
(-3.001)– 9.0100*** –
Market model 1.2300a
(136.188)0.0000(0.055)
– 0.0051(1.1908)
– 1.4180
Inchcape—pre-crisis period
Joint model 0.9864a
(91.873)0.0000(0.617)
0.0188(1.4516)
-0.1231a
(-5.8971)– –
Idiosyncratic model 1.002a
(94.820)0.0000(0.4900)
-0.0160(-1.1948)
– 1.4277 –
T. Choudhry, R. Jayasekera
123
applicable and it indicates market efficiency with a negative market shock of small size.
During the crisis period there is some evidence of asset mispricing via the non-market shock.
The coefficient on the non-market shock is positive and significant but also small in size. As
stated earlier, asset mispricing can be in the form of over/under-reaction to information
which essentially suggests market inefficiency. Barclays Bank and Standard Charter also
provide similar results with more evidence of market efficiency than asset mispricing.
Results from the retail industry are shown in Table 6. In general, both periods tend to
provide support for the market efficiency. The model selection does not require a change of
model from pre-crisis to crisis period for all firms except NEXT. Furthermore, in the case
of Inchcape, using the market model in both periods, the coefficients on the market shock
(dm) are negative and significant. There is a decline in the size of the coefficient from the
pre-crisis to crisis period indicating a declining market efficiency. Similar results are
provided by the Home Retail Group via the idiosyncratic model. NEXT also shows similar
results with the market model during the pre-crisis period and the idiosyncratic model
during the crisis period. Declining market efficiency is a clear indication of the affect of the
crisis. Only Kingfisher and Marks and Spencer show some diversity in results between the
two periods. They provide some evidence of asset mispricing. For example in the case of
Marks and Spencer, based on the idiosyncratic model, the non-market shock changes from
a negative to a positive between the two periods. The size of the effect in absolute value is
similar.
Table 6 continued
Company a ab di dm H0: di = 0 H0: dm = 0
Market model 0.9810a
(96.7686)0.0000(0.6370)
– -0.1061a
(-6.1314)– 37.5930***
Inchcape—crisis period
Joint model 1.0978a
(73.257)-0.0000c
(-1.6980)0.0125b
(2.1653)-0.0549a
(-4.0726)– –
Idiosyncratic model 1.1195a
(71.2055)-0.0000(-1.4504)
0.0033(0.4881)
– 0.2383 –
Market model 1.0884a
(75.4039)-0.0000c
(-1.7375)– -0.0400a
(-3.3914)– 11.5018***
Home retail group—pre-crisis period
Joint model 0.9071a
(99.1432)0.0000(0.6892)
-0.0248(-1.5909)
0.0070(0.3506)
– –
Idiosyncratic model 0.9064a
(100.379)0.0000(0.6623)
-0.0219b
(-2.001)– 4.0030** –
Market model 0.9123a
(101.385)0.0000(0.5215)
– -0.0142(-0.9845)
– 0.9693
Home retail group—crisis period
Joint model 1.1462a
(106.793)-0.0000(-0.919)
-0.0244a
(-3.5112)0.0260a
(3.2774)– –
Idiosyncratic model 1.1405a
(108.704)-0.0000(-0.9237)
-0.0148a
(-2.8776)– 8.2808*** –
Market model 1.1600a
(110.473)-0.0000(-1.0600)
– 0.0031(0.5322)
– 0.2832
t-statistics in the parenthesesa ,b, and c Significantly different from zero at the 1, 5 and 10 % level, respectively
***, ** and * Rejection of the null hypothesis at the 1, 5 and 10 % level, respectively
Level of efficiency in the UK equity market
123
Table 7 Estimation of the Cho and Engle models (industry = oil)
Company a ab di dm H0: di = 0 H0: dm = 0
British petroleum—pre-crisis period
Joint model 1.053a
(129.290)0.0000(0.2615)
-0.0554a
(-4.0021)0.0664a
(3.9152)– –
Idiosyncratic model 1.049a
(128.287)0.0000(0.3894)
-0.0257a
(-2.7702)– 7.6741*** –
Market model 1.065a
(136.873)0.0000(0.2286)
– 0.0146(1.2146)
– 1.4751
British petroleum—crisis period
Joint model 0.9126a
(151.650)0.0000(0.0303)
0.0052(0.6672)
-0.0168b
(-1.9954)– –
Idiosyncratic model 0.9158a
(149.322)0.0000(0.1326)
-0.0041(-0.8607)
– 0.7408 –
Market model 0.9114a
(148.685)0.0000(0.0577)
– -0.0120b
(-3.2289)– 10.4256***
Royal Dutch—pre-crisis period
Joint model 1.0581a
(154.770)-0.0000(-0.5463)
-0.0171c
(-1.7684)0.0344a
(2.7367)– –
Idiosyncratic model 1.0550a
(158.490)-0.0000(-0.5154)
-0.0026(-0.4099)
– 0.1680 –
Market model 1.0612a
(160.877)-0.0000(-0.5784)
– 0.0180c
(1.9475)– 3.7926*
Royal Dutch—crisis period
Joint model 0.9344a
(155.662)-0.0000(-1.2674)
-0.0147b
(-2.3757)0.0113(1.5611)
– –
Idiosyncratic model 0.9333a
(156.208)-0.0000(-1.3300)
-0.0067a
(-2.7000)– 7.2889*** –
Market model 0.9367a
(157.052)-0.0000(-1.3713)
– -0.0028(-0.8886)
– 0.7896
BG group—pre-crisis period
Joint model 1.1137a
(113.272)-0.0000(-1.4266)
0.0346a
(3.0231)-0.0243(-1.4117)
– –
Idiosyncratic model 1.1159a
(114.872)-0.0000(-1.4312)
0.0264a
(3.1450)– 9.8909*** –
Market model 1.1042a
(117.368)-0.0000(-1.4741)
– 0.0107(0.8531)
– 0.7278
BG group—crisis period
Joint model 1.0841a
(189.978)-0.0000(-0.174)
0.0205a
(4.7132)-0.0231a
(-3.8321)– –
Idiosyncratic model 1.0866a
(190.334)-0.0000(-0.098)
0.0082a
(-3.2685)– 10.6832*** –
Market model 1.077a
(192.118)-0.0000(-0.0802)
– -0.0014(-0.4371)
– 0.1910
Tullow—pre-crisis period
Joint model 1.2202a
(75.280)0.0000c
(1.732)-0.0508a
(-3.8331)0.0359(1.1922)
– –
Idiosyncratic model 1.2163a
(76.000)0.0000c
(1.775)-0.0425a
(-4.1652)– 17.3490*** –
T. Choudhry, R. Jayasekera
123
The results of the oil industry firms (Table 7) provide mixed results. Results provide
evidence supportive of both market efficiency and asset mispricing. British Petroleum is
the only firm that shows evidence of market efficiency during both periods. BG Group and
ENSCO provide evidence of asset mispricing during both periods via the idiosyncratic
mode, although there is indication of a declining asset mispricing. Oil prices can be
difficult to predict thus making it hard for the investors to gauge the risk level. This may
lead to asset mispricing due to the difficulty in determining the premium to compensate for
the estimation of risk resulting from the uncertainty.26 The model selection fails to select
the joint model for any of the firms during any of the two periods.
The results of the food industry firms (Table 8) indicate conclusions similar to the
banking and retail industries. The results from this industry though, do provide support for
the market efficiency and asset mispricing during both periods, especially during the crisis
period. Again the sizes of the coefficient on the market shock (dm) and non-market shock
(di) are smaller during the crisis period, implying declining market efficiency and asset
mispricing. This is clearly visible in the case of Associated British Food where the model
Table 7 continued
Company a ab di dm H0: di = 0 H0: dm = 0
Market model 1.2400a
(80.514)0.0000c
(1.770)– -0.0309
(-1.1882)– 1.4118
Tullow—crisis period
Joint model 1.0875a
(116.513)0.0000(1.1406)
-0.0028(-0.4268)
0.0108(1.1544)
– –
Idiosyncratic model 1.0858a
(116.080)0.0000(1.1833)
0.0019(0.5555)
– 0.3086 –
Market model 1.0887a
(117.170)0.0000(1.1502)
– 0.0078c
(1.7459)– 3.0481*
ENSCO—pre-crisis period
Joint model 0.6482a
(35.0147)-0.0000b
(-2.1193)0.0616a
(6.0381)-0.0504b
(-2.0694)– –
Idiosyncratic model 0.6586a
(37.7374)-0.0000b
(-2.0647)0.0571a
(5.7372)– 32.9160*** –
Market model 0.6045a
(36.141)-0.0000b
(-2.2076)– -0.0122
(-0.4550)– 0.2070
ENSCO—crisis period
Joint model 0.9602a
(90.280)0.0000(0.300)
0.0172a
(4.3981)-0.0163b
(-2.1751)– –
Idiosyncratic model 0.9654a
(92.383)0.0000(0.2399)
0.0130a
(4.4767)– 20.0410*** –
Market model 0.9490a
(92.4439)0.0000(0.2480)
– 0.0003(0.0549)
– 0.0030
t-statistics in the parenthesesa ,b, and c Significantly different from zero at the 1, 5 and 10 % level, respectively
***, ** and * Rejection of the null hypothesis at the 1, 5 and 10 % level, respectively
26 ‘‘Appendix’’ shows how repeated attempts at predicting the oil price made by the US Department ofEnergy exhibited huge deviations from the actual price levels, thus serving to illustrate the difficulty inpredicting the movements in the oil prices.
Level of efficiency in the UK equity market
123
Table 8 Estimation of the Cho and Engle models (industry = food)
Company a ab di dm H0: di = 0 H0: dm = 0
TESCO—pre-crisis period
Joint model 0.6022a
(61.491)0.0000(0.3386)
0.1417a
(6.2400)-0.2135a
(-9.9630)– –
Idiosyncratic model 0.6295a
(57.218)0.0000(0.1765)
0.0663a
(2.9632)– 8.7800*** –
Market model 0.5675a
(59.996)0.0000(0.4687)
– -0.1224a
(-6.4621)– 41.7580***
TESCO—crisis period
Joint model 0.8496a
(130.255)0.0000(0.0640)
0.0413a
(4.9658)-0.0527a
(-7.2737)– –
Idiosyncratic model 0.8587a
(129.148)0.0000(0.2549)
0.0075(1.3316)
– 1.7732 –
Market model 0.8383a
(129.681)0.0000(0.5379)
– -0.0229a
(-4.5405)– 20.6163***
Sainsbury—pre-crisis period
Joint model 0.9046a
(71.881)0.0000(0.3863)
-0.1297a
(-7.7274)0.1740a
(8.3600)– –
Idiosyncratic model 0.8815a
(68.447)0.0000(0.1144)
-0.0716a
(-4.0612)– 16.4934*** –
Market model 0.9318a
(77.204)-0.0000(-0.2042)
– 0.0680a
(3.0281)– 9.1700***
Sainsbury—crisis period
Joint model 0.9016a
(63.178)-0.0000b
(-2.4187)-0.1116a
(-3.9790)0.1039a
(4.4620)– –
Idiosyncratic model 0.8714a
(39.330)-0.0000b
(-2.1472)-0.0668b
(-2.4469)– 5.9873** –
Market model 0.9336a
(70.2629)-0.0000b
(-2.0186)– 0.0099
(1.2757)– 1.6270
Morrison—pre-crisis period
Joint model 0.6541a
(48.894)0.0000(0.6237)
0.0853a
(2.8683)-0.1859a
(-6.6402)– –
Idiosyncratic model 0.6874a
(39.246)0.0000(0.6042)
0.0502(1.4330)
– 2.0536 –
Market model 0.6226a
(61.575)0.0000(1.1366)
– -0.1355a
(-7.8715)– 61.9602***
Morrison—crisis period
Joint model 0.8549a
(167.377)-0.0000(-1.5259)
-0.0196a
(-4.2713)-0.0175a
(-3.9055)– –
Idiosyncratic model 0.8582a
(173.752)-0.0000c
(-1.6600)-0.0305a
(-7.1929)– 51.7371*** –
Market model 0.8603a
(179.478)0.0000(-1.2782)
– -0.0312a
(-6.8303)– 46.6530***
Associate british food—pre-crisis period
Joint model 0.5570a
(50.833)-0.0000(-0.4346)
0.1912a
(4.0431)-0.1928a
(-5.6200)– –
Idiosyncratic model 0.5826a
(36.7359)0.0000(0.1421)
0.1130b
(2.4705)– 6.1033*** –
T. Choudhry, R. Jayasekera
123
selection picks the joint model during both periods. The positive coefficient on the non-
market shock and the negative coefficient on market shock declines from the pre to the
crisis period without changing the direction of the affect on the beta. Similar analysis can
be extended to the other firms even though different models are employed during the two
periods for each firm. Only Unilever provides somewhat of a unique result. The market
model during the pre-crisis period provides evidence of asset mispricing in the form of a
significant positive coefficient on the market shock. During the crisis period, the coefficient
on the market shock in the joint model is still positive at the same size, however now the
non-market shock is negative indicating some evidence of market efficiency.
The construction and land industry (Table 9) provides results in line with other
industries (except oil). For all firms except Keller different models are selected during both
periods. Both periods support the market efficiency and the declining size of the coeffi-
cients from the pre-crisis to crisis period indicates declining market efficiency. The small
evidence of asset mispricing is quite weak.
So what do our results show? Most firms and industries seem to support the market
efficiency during both periods. The exception to this being the oil industry, though this may
be due to the specific uncertainty surrounding the future oil price determination. The level
of market efficiency however seems to decline significantly from the pre-crisis to the crisis
Table 8 continued
Company a ab di dm H0: di = 0 H0: dm = 0
Market model 0.5209a
(50.2191)-0.0000(-0.1072)
– -0.0747a
(-4.0529)– 16.4260***
Association British food—crisis period
Joint model 0.8112a
(121.893)-0.0000(-0.0826)
0.0642a
(6.2226)-0.0576a
(-7.5683)– –
Idiosyncratic model 0.8221a
(121.002)0.0000(0.0075)
0.0219a
(3.2353)– 10.4670*** –
Market model 0.8006a
(120.628)0.0000(0.1500)
– -0.0117b
(-2.5171)– 6.3360**
Unilever—pre-crisis period
Joint model 0.8126a
(99.957)-0.0000(-0.0729)
-0.0474b
(-2.3100)0.0829a
(3.9234)– –
Idiosyncratic model 0.8019a
(93.917)0.0000(0.1215)
-0.0169(-1.0503)
– 1.1030 –
Market model 0.8223a
(105.460)-0.0000(-0.0214)
– 0.0487a
(2.9128)– 8.4830***
Unilever—crisis period
Joint model 0.7896a
(98.479)-0.0000(-0.6541)
-0.0451a
(-4.7185)0.0487a
(4.8338)– –
Idiosyncratic model 0.7798a
(91.3887)-0.0000(-0.7266)
-0.0145a
(-3.0529)– 9.3200*** –
Market model 0.8014a
(104.834)-0.0000(-0.6931)
– 0.0171a
(2.704)– 7.3129***
t-statistics in the parenthesesa ,b, and c Significantly different from zero at the 1, 5 and 10 % level, respectively
***, ** and * Rejection of the null hypothesis at the 1, 5 and 10 % level, respectively
Level of efficiency in the UK equity market
123
Table 9 Estimation of the Cho and Engle models (industry = construction and land)
Company a ab di dm H0: di = 0 H0: dm = 0
Keller—pre-crisis period
Joint model 0.4560a
(23.844)-0.0000(-0.5373)
0.0224(1.0329)
-0.2226a
(-7.4935)– –
Idiosyncratic model 0.4982a
(27.634)-0.0000(-0.2480)
-0.0165(-0.6665)
– 0.4442 –
Market model 0.4482a
(26.219)-0.0000(-0.5182)
– -0.2006a
(-6.7683)– 45.8100***
Keller—crisis period
Joint model 1.0120a
(61.882)-0.0000(-0.4039)
0.0286b
(2.3413)-0.0868a
(-5.2749)– –
Idiosyncratic model 1.040a
(57.991)-0.0000(-0.3119)
0.0027(0.2074)
– 0.0430 –
Market model 0.9947a
(63.958)-0.0000(-0.2285)
– -0.0635a
(-5.6392)– 31.8007***
Kier—pre-crisis period
Joint model 0.8499a
(58.456)-0.0000(-0.4207)
-0.0632a
(-3.4516)-0.0070(-0.2403)
– –
Idiosyncratic model 0.8510a
(59.294)-0.0000(-0.4294)
-0.0649a
(-4.5247)– 20.4730*** –
Market model 0.8663a
(60.782)-0.0000(-0.1584)
– -0.0683a
(-2.9511)– 8.7093***
Kier—crisis period
Joint model 0.9713a
(84.459)-0.0000(-0.6011)
-0.0215a
(-4.0746)0.0132c
(1.8116)– –
Idiosyncratic model 0.9678a
(85.704)-0.0000(-0.6035)
-0.0169a
(-3.9754)– 15.8040*** –
Market model 0.9840a
(90.152)-0.0000(-0.6783)
– -0.0094(-0.6612)
– 0.4371
James Halstead—pre-crisis period
Joint model 0.2160a
(18.080)-0.0000(-0.6185)
0.0274(1.4927)
-0.0978a
(-5.0644)– –
Idiosyncratic model 0.2421a
(22.513)-0.0000(-0.7500)
0.0095(0.4526)
– 0.2099 –
Market model 0.2111a
(18.1800)-0.0000(-0.6120)
– -0.0884a
(-4.9327)– 24.3310***
James Halstead—crisis period
Joint model 0.4982a
(43.287)-0.0000(-0.5346)
-0.0158b
(-2.0177)-0.0306a
(-4.0970)– –
Idiosyncratic model 0.5127a
(47.768)-0.0000(-0.7155)
-0.0264a
(-3.4232)– 11.7184*** –
Market model 0.5039a
(44.621)-0.0000(-0.561)
– -0.0367a
(-5.2736)– 27.8110***
SEGRO—pre-crisis period
Joint model 0.8600a
(76.612)-0.0000(-0.2819)
-0.0097(-0.3809)
-0.0754a
(-2.6745)– –
Idiosyncratic model 0.8684a
(73.188)-0.0000(-0.1990)
-0.0372b
(-1.9726)– 3.8911** –
T. Choudhry, R. Jayasekera
123
period. Our findings show that this decline is most prominent and unambiguous with the
banking industry closely followed by the construction and land industry. These effects
appear to be less pronounced with the retail industry and even less with regards to food,27
although these effects are still prevalent. Both results of market efficiency and declining
market efficiency from the pre-crisis to crisis periods provide ample evidence of the affect
of the current financial crisis on beta.
We present circumstantial evidence by citing the behaviour of the VSTOXX28 Index,
which can be regarded as a proxy for the volatility in the Eurozone from pre-crisis to crisis
period, which strengthens our argument. In general, the volatility in the Eurozone increased
in excess of 226 % during the crisis period (Bloomberg). The average Eurozone volatility
increased from 11.089 from the pre-crisis period to 25.107 during the crisis period. This
behaviour can be interpreted as a decline in the level of market efficiency as this excess
volatility can be attributed to asset prices over/under shooting (i.e. asset mispricing) rel-
ative to their intrinsic values. This in effect would cause a series of repetitive corrections to
Table 9 continued
Company a ab di dm H0: di = 0 H0: dm = 0
Market model 0.8621a
(81.186)-0.0000(-0.3162)
– -0.0842a
(-5.2800)– 27.8760***
SEGRO—crisis period
Joint model 1.1400a
(113.024)-0.0000(-1.2444)
-0.0007(-0.1538)
0.0122(1.5206)
– –
Idiosyncratic model 1.1356a
(115.977)-0.0000(-1.3155)
0.0025(0.6986)
– 0.4880 –
Market model 1.1400a
(116.137)-0.0000(-1.2432)
– 0.0115c
(1.8460)– 3.4077*
Capital shopping centre group—pre-crisis period
Joint model 0.9556a
(98.402)0.0000(0.0241)
-0.1039a
(-6.6382)0.0527b
(2.2365)– –
Idiosyncratic model 0.9507a
(99.2301)0.0000(0.1304)
-0.0801a
(-8.3980)– 70.5261*** –
Market model 0.9682a
(100.142)0.0000(0.6533)
– -0.0585a
(-3.4813)– 12.1194***
Capital shopping centre group– crisis period
Joint model 1.0375a
(122.059)-0.0000(-0.6700)
-0.0120b
(-2.4200)0.0104c
(1.7079)– –
Idiosyncratic model 1.0357a
(123.305)-0.0000(-0.7205)
-0.0074b
(-2.1078)– 4.4430** –
Market model 1.0427a
(127.661)-0.0000(-0.6801)
– -0.0015(-0.3365)
– 0.1132
t-statistics in the parenthesesa ,b, and c Significantly different from zero at the 1, 5 and 10 % level, respectively
***, ** and * Rejection of the null hypothesis at the 1, 5 and 10 % level, respectively
27 This is intuitive as the demand for food is relatively inelastic.28 VSTOXX Index, developed by Deutsche Borse and Goldman Sachs is a measure of volatility in theEurozone. It measures implied volatility on options across all maturities.
Level of efficiency in the UK equity market
123
achieve convergence to their intrinsic values in a dynamic sense, thus in theory presenting
more profitable arbitrage opportunities especially for speculators and hedge funds.
Our results have interesting implications to investors, especially those working with
hedge funds. We find that notions of market efficiency hold during the pre-crisis and the
crisis period, however the level of this efficiency declines significantly during the crisis
period. If the markets are efficient this would suggest that the asset prices would converge
almost instantaneously to their intrinsic values, thus wiping off any opportunities of prof-
iting through arbitrage. We find that the level of market efficiency significantly declines as
the UK economy slides into a recession during the crisis period, suggesting a delay in the
convergence of asset prices to their intrinsic values, thus in theory opening up more arbi-
trage opportunities. In a frictionless market value-relevant information is instantaneously
incorporated into market prices. Such a market is deemed to be informational efficient.
Arbitrage is the mechanism that disciplines market prices and achieves this efficiency. In the
absence of trading frictions, costless arbitrage guides the market price discovery towards the
‘‘fundamental values’’ at least in the medium term. Trading costs introduce constrained
arbitrage pressure and the incorporation of information into market prices can be sub-
stantially delayed. Lesmond et al. (2004) conclude that the delay in price adjustment for
security returns simply reflects the costs of arbitrage creating an illusion of anomalous price
behaviour and momentum trading profit opportunity when, in fact, none exists. Their evi-
dence suggests that momentum patterns are largely an artefact of the slow price updating of
high transaction cost stocks. The model we apply assesses the level of market efficiency
based on the share price fluctuations and the speed of the price adjustment process. We find
that the level of efficiency of the UK equity market has dropped from the pre-crisis to the
crisis period, and this may be due to the presence of trading friction in the form of trans-
action costs as suggested by Lesmond et al. (2004). However, we do not instigate the level
of changes in the transaction costs in the UK during the transition from the pre-crisis to
crisis period here and hence do not wish to attribute the decline in the level of market
efficiency purely to this reason. Neither do we study the actual quantum of the arbitrage
profits that can be realised in order to comment on the ‘‘illusory’’ nature of these profits as
claimed by Lesmond et al.(2004), as this is be beyond the scope of this paper. However we
believe that establishing to what extent these theoretical profits can be realised, especially in
the UK equity markets during the crisis period, is an important area that warrants future
research.
8 Conclusion
The controversy surrounding the abnormality of stock prices has been a subject of
extensive research over the past few decades. The controversy can essentially be viewed as
two competing mutually independent hypotheses explaining certain aspects of stock price
behaviour. The first, ‘asset mispricing’, puts forth a behavioural finance argument to
explain certain anomalies of stock price behaviour. The alternative argument, for ‘market
efficiency’ serves to enforce the efficient market hypothesis and provide evidence that the
beta of individual stock rises (falls) in response to abnormally negative (positive) returns,
and argues that this asymmetric response to good and bad news explains the performance
of stock returns.
Given the evidence on the predictive asymmetry of volatility, we investigate the
asymmetric effect of beta during good and bad periods to good and bad news and shed
fresh insight on the controversy of the ‘‘abnormalities of stock prices’’. i.e. whether the
T. Choudhry, R. Jayasekera
123
hypothesis of ‘increased market efficiency’ or ‘asset mispricing’ would better fit the
empirical observations as an economy slides from a boom to a recession. We focus on
the UK equity market, analysing the relative affects on five major industrial segments
using a bivariate BEKK GARCH approach. This framework is used to capture the affects
of a period when the UK economy slid from relative prosperity (pre-crisis) to a reces-
sion, i.e. the current financial crisis period. We define the pre-crisis period (which we
refer to as the ‘‘good’’ period) from January 2004 to June 2007 and the crisis period
(which we refer to as the ‘‘bad’’ period) to commence from June 2007 to September
2010. Thus, we investigate seven years of daily data, classified under 5 major industries
covering 25 companies in total. The industries under study are; Banking, Retail, Food,
Construction and Oil.
So what do our results show? The GARCH results are quite standard. We document,
for the first time using UK equity market data, the level of market efficiency as the
economy slides from a relative boom to a recession. We find that most firms and
industries seem to support the market efficiency hypothesis during both periods. The
exception to this being the oil industry, which result may be due to the uncertainty
surrounding the future oil price determination. The level of market efficiency, however,
seems to decline significantly from the pre-crisis to crisis period. Some exceptions are
provided by a few firms in the retail industry, perhaps due to the inelastic demand and
less uncertainty surrounding the determination of factor costs. These results have
implications to investors, especially hedge funds, as essentially they would be presented
at least in theory with relatively more profitable arbitrage opportunities during a crisis
period.
As pointed out by Malkiel (2003) pricing irregularities and even predictable patterns in
stock returns can appear over time and even persist for short periods. Markets cannot be
perfectly efficient, or there would be no incentive for professionals to uncover the infor-
mation that gets so quickly reflected in market prices, a point stressed by Grossman and
Stiglitz (1980). In this context our study is a relatively short term study. It is quite plausible
that the markets may behave less efficiently during certain periods, in the short term at
least. As Fama (1998) concludes, the market efficiency hypothesis offers a simple answer
to this type of inefficient period where the market prices stray from the fundamental values.
Specifically, Fama (1998) claims that the expected value of abnormal returns is zero, but
chance generates apparent anomalies that split randomly between overreaction and under-
reaction. On the other hand, consistently generating arbitrage profits by exploiting these
anomalies may not be straightforward. As Timmermann and Granger (2004) claim, stable
forecasting patterns are unlikely to persist for long periods of time and will self-destruct
when discovered by a large number of investors. Both results of market efficiency and
declining market efficiency from the pre-crisis to crisis periods provide ample evidence of
the asymmetric effect of the current financial crisis on beta. Given the status of the current
crisis our results advocate future research in this field using data from different countries,
different firms and using different methods.
Acknowledgments We thank an anonymous referee for several useful comments. Any remaining errorsare the authors’ responsibility.
Appendix
See Fig. 2.
Level of efficiency in the UK equity market
123
References
Agrawal A, Jaffe F, Mandelker G (1992) The post-merger performance of acquiring firms: a re-examinationof an anomaly. J Financ 47:1605–1621
Asquith P (1983) Merger bids, uncertainty and stockholder returns. J Financ Econ 11:51–83Ball R, Brown P (1968) An empirical evaluation of accounting income numbers. J Account Res 6:159–178Ball R, Kothari SP (1989) Nonstationary expected returns: implications for tests of market efficiency and
serial and correlation in returns. J Financ Econ 25:51–74Bartram SM, Bodnar GM (2009) No place to hide: the global crisis in equity markets in 2008/2009. J Int
Money Financ 28:1246–1292Bernard V, Thomas J (1990) Evidence that stock prices do not fully reflect the implications of current
earnings for future earnings. J Account Econ 13:305–322Black F (1976) Studies of stock market volatility changes. In: Proceedings of the American Statistical
Association. Business and Economics Statistics Section, pp 177–181Bodurtha J, Mark N (1991) Testing the CAPM with time-varying risk and returns. J Financ 46:1485–1505Bollerslev T, Engle RF, Wooldridge JM (1988) A capital asset pricing model with time varying covariances.
J Polit Econ 96:116–131Bollerslev T, Chou R, Kroner K (1992) ARCH modeling in finance. J Econom 52:5–59Braun P, Nelson D, Sunier A (1995) Good news, bad news, volatility and betas. J Financ 50:1575–1601Brooks C, Henry O (2002) The impact of news on measures of undiversifiable risk: evidence from the UK
stock market. Oxf Bull Econ Stat 64:487–507Campbell J, Hentschel L (1992) No news is good news: an asymmetric model of changing volatility in stock
return. J Financ Econ 31:281–318Chan KC (1988) On the contrarian investment strategy. J Bus 61:147–163Cho YH, Engle R (1999) Time-varying betas and symmetric effects of news: empirical analysis of blue chip
stocks. National Bureau of Economic Research. Working paper no. 7330Chopra N, Lakinishok J, Ritter J (1992) Measuring abnormal returns: do stocks overreact? J Financ Econ
10:289–321Choudhry T (2002) Stochastic Structure of the time-varying beta: evidence from the UK companies. Manch
Sch 70:768–791Choudhry T, Jayasekera R (2012) Comparison of efficiency characteristics between the banking sectors of
US and UK during the global financial crisis of 2007–2011. Int Rev Financ Anal 25:106–116Christie A (1982) The Stochastic behavior of common stock variances: value, leverage and interest rate
effects. J Financ Econ 10:407–432Cusatis P, Miles J, Wooldridge J (1993) Restructuring through spinoffs. J Financ Econ 33:293–311De Bondt WFH, Thaler R (1989) Anomalies: a mean-reverting walk down wall street. J Econ Perspect
3:189–202Desai H, Jain P (1997) Long-run common stock returns following splits and reverse splits. J Bus 70:409–433
Fig. 2 US DOE oil price forecasts. Source: U.S. Department of Energy, 1998
T. Choudhry, R. Jayasekera
123
Dharan B, Ikenberry D (1995) The long-run negative drift of post-listing stock returns. J Financ50:1547–1574
Dias JG, Ramos SB (2013) The aftermath of the subprime crisis: a clustering analysis of the world bankingsector. Forthcoming Review of Quantitative Accounting and Finance
Dwyer GP, Lothian JR (2012) International and historical dimensions of the financial crisis of 2007 and2008. J Int Money Financ 31:1–9
Dwyer GP, Tkac P (2009) The financial crisis of 2008 in fixed-income markets. J Int Money Financ28:1293–1316
Engle R, Kroner K (1995) Multivariate simultaneous generalized ARCH. Econom Theory 11:122–150Engle R, Ng VM, Rothschild M (1990) Asset pricing with a factor-ARCH structure: empirical estimates for
treasury bills. J Econom 45:213–237European Commission (2009) Economic crisis in Europe: causes, consequences and responses. European
EconomyFama E (1970) Efficient capital markets: a review of theory and empirical work. J Financ 25:383–417Fama E (1991) Efficient capital markets: II. J Financ 46:1575–1617Fama E (1998) Market efficiency, long term returns, and behavioral finance. J Financ Econ 49:283–306Fama E, French K (1992) The cross-section of expected stock returns. J Financ 47:427–465Fama E, French K (1993) Common risk factors in the returns on stocks and bonds. J Financ Econ 33:3–56Fama E, French K (1998) Value versus growth: the international evidence. J Financ 53:1975–1999Fama E, French K (2002) Testing trade-off and pecking order predictions about dividends and debt. Rev
Financ Stud 15:1–33Frazzini A (2006) The Disposition effect and underreaction to news. J Financ 61:2017–2046French KR, Schwert GW, Stambaugh RF (1987) Expected stock returns and volatility. J Polit Econ
99:385–415Giannopoulos K (1995) Estimating the time-varying components of international stock markets risk. Eur J
Financ 1:129–164Grossman SJ, Stiglitz JE (1980) On the impossibility of informationally efficient markets. Am Econ Rev
70:393–408Hansen L, Richard S (1987) The role of conditioning information in deducing testable restriction implied by
dynamic asset pricing models. Econometrica 55:587–614Harel A, Harpaz G, Francis JC (2011) Analysis of efficient markets. Rev Quant Financ Acc 36:287–296Ikenberry D, Lakonishok J (1993) Corporate governance through the proxy contest: evidence and impli-
cations. J Bus 66:405–435Ikenberry D, Lakonishok J, Vermaelen T (1995) Market underreaction to open market share repurchases.
J Financ Econ 39:181–208Ikenberry D, Rankine G, Stice E (1996) What do stock splits really signal? J Financ Quant Anal 31:357–377Jegadeesh N, Titman S (1993) Returns to buying winners and selling losers: implications for stock market
efficiency. J Financ 48:65–91Kamin SB, DeMarco LP (2012) How did a domestic housing slump turn into a global financial crisis? J Int
Money Financ 31:10–41Klemkosky R, Martin J (1975) The adjustment of beta forecasts. J Financ 30:1123–1128Lakonishok J, Vermaelen T (1990) Anomalous price behaviour around repurchase tender offers. J Financ
45:455–477Lesmond DA, Schillb MJ, Zhouc C (2004) The illusory nature of momentum profits. J Financ Econ
71:349–380Lintner J (1965) The valuation of risk assets and the selection of risky investments in stock portfolios and
capital budgets. Rev Econ Stat 47:13–37Loughran T, Ritter J (1995) The new issues puzzle. J Financ 50:23–51Malkiel GB (2003) The efficient market hypothesis and Its critics. J Econ Perspect 17:59–82Markowitz H (1952) Portfolio selection. J Financ 7:77–91Marsh T, Pfleiderer P (2012) Black swans and the financial crisis. Rev Pac Basin Financ Markets Policies
15:1–12Michaely R, Thaler R, Womack K (1995) Price reactions to dividend initiations and omissions. J Financ
50:573–608Muth J (1961) Rational expectation and the theory of price movements. Econometrica 29:1–23Nelson DB (1991) Conditional heteroskedasticity in asset returns: a new approach. Econometrica
59:347–370Pindyck RS (1984) Risk, inflation, and the stock market. Am Econ Rev 74:335–351Poterba TM, Summers LH (1986) The persistence of volatility and stock market fluctuations. Am Econ Rev
76:1142–1151
Level of efficiency in the UK equity market
123
Ritter J (1991) The long-term performance of initial public offerings. J Financ 46:3–27Roll R (1986) The hubris hypothesis of corporate takeovers. J Bus 59:197–216Schwert GW (1989) Why does stock market volatility change over time. J Financ 44:1115–1154Sentance A, Taylor MP, Wieladek T (2012) How the UK economy weathered the financial storm. J Int
Money Financ 31:102–123Sharpe W (1964) Capital asset prices: a theory of market equilibrium under conditions of risk. J Financ
19:425–442Shin H (2009) Reflections on Northern Rock: the bank run that heralded the global financial crisis. J Econ
Perspect 23:101–119Spiess D, Affleck-Graves J (1995) Underperformance in long-run stock returns following seasoned equity
offerings. J Financ Econ 38:243–267Susmel R, Engle R (1994) Hourly volatility spillovers between international equity markets. J Int Money
Financ 13:3–25Timmermann A, Granger CWJ (2004) Efficient market hypothesis and forecasting. Int J Forecast 20:15–27Veronesi P (1999) Stock market overreaction to bad news in good times: a rational expectations equilibrium
model. Rev Financ Stud 12:975–1007
T. Choudhry, R. Jayasekera
123