Technical efficiency in the English Football Association Premier League with a stochastic cost frontier

Page created by Angel Miranda
 
CONTINUE READING
Applied Economics Letters, 2007, 14, 731–741

Technical efficiency in the English
Football Association Premier League
with a stochastic cost frontier
Carlos Pestana Barrosa,* and Stephanie Leachb
a
 Instituto Superior de Economia e Gestão, Technical University of Lisbon,
Rua Miguel Lupi, 20 1249-078 Lisbon, Portugal
b
  Tanaka Business School, Imperial College, London

This article uses an econometric frontier model to evaluate the technical
efficiency of English Premier League clubs from 1998/99 to 2002/03
combining sport and financial variables. A Cobb–Douglas cost specifica-
tion of the technical efficiency effects model is used to generate football
club efficiency scores, allowing for contextual variables which affect
inefficiency. We conclude that the efficiency scores are mixed. A policy
is devised for the management of this sector.

I. Introduction                                                        The motivation for the present article is derived
                                                                    from stylised facts observed in the English football
In this article, we measure the technical efficiency                industry, such as the clubs which overspend in order
of the clubs playing in the English Premier League                  to achieve sporting success, but then fail to do so,
with a Cobb–Douglas cost frontier model, using                      e.g. Leeds United. In this case, the failure may be due
data obtained in the Deloitte & Touche reports on                   to uneven playing fields in the Premier League, in
English football from 1998/99 to 2002/03. Previous                  which the market leaders in terms of turnover appear
research into the English Premier League has made                   to be virtually guaranteed sporting success. In this
use of data envelopment analysis (DEA), e.g. Haas                   case, the clubs playing in sub-championships of their
(2003b) and of the stochastic frontier model, Dawson                own, and with very different objectives from the few
et al. (2000). However, none of the articles adopted                elite clubs, sometimes start overspending in an
the stochastic frontier model with contextual vari-                 attempt to achieve the elite position, but usually
ables, known as the technical efficiency effects model              fail to do so. Alternatively, failure may be due to
(Coelli et al., 1998).                                              technical inefficiency, since when a club starts over-
   This article analyses the efficiency of the English              spending it expects to overcome the uneven status
Premier League with the use of a technical efficiency               quo, but may lack the managerial skills to do so.
model, allowing for contextual variables which may                  Additional reasons are exogenous contextual effects,
affect the clubs’ performance. In fact, based on the                such as the population and income of the club’s fan
sports economics literature (El-Hodiri and Quirk,                   base, which defines an inescapable environment that
1971; Fort and Quirk, 1995) it is expected that the                 condemns clubs with small bases to life outside
clubs’ fan base will have a major effect on their                   the top. Finally, there are exogenous shocks such
performance. As the club base is composed of the                    as the Abramovich effect, presently observed at
population and income in the club area, these                       Chelsea, which translates into changes in the relative
contextual variables have to be included in the cost                efficiency of the clubs in a league, circumventing
function.                                                           the club base.

*Corresponding author. E-mail: cbarros@iseg.utl.pt
                   Applied Economics Letters ISSN 1350–4851 print/ISSN 1466–4291 online ß 2007 Taylor & Francis         731
                                                   http://www.tandf.co.uk/journals
                                                  DOI: 10.1080/13504850600592440
732                                                                                   C. P. Barros and S. Leach
   This article extends previous research into              Premiership teams spending heavily and incurring
football efficiency, adopting a stochastic frontier         operating losses in the hope of achieving a certain
model, alongside Hoeffler and Payne (1997) and              position guaranteeing qualification for European
Dawson et al. (2000) to evaluate the technical              competition.
efficiency of the English Premier League football              In order to compete for playing talent with other
clubs. However, this article adopts the technical           major teams in Europe (e.g. Real Madrid and
efficiency effects model, found in Coelli et al.            Juventus), English teams have had to increase their
(1998), which allows for contextual variables in the        spending on wages in order to attract the best players.
cost function. A sample of the clubs that played            Furthermore, the biggest teams in England, notably
consecutively in the league in the years under analysis     Manchester United, Arsenal, Liverpool and more
(1998/99 to 2002/03) is used. The use of such clubs         recently Chelsea, have increasingly imported players
ensures balanced panel data and is needed                   (and in the case of the latter three, managers) from
to obtain similar average scores over the period at         ‘overseas’ with the result that perhaps only a handful
club level.                                                 of English players have a place in the starting 11.
   The article is organized as follows: in Section II, we   Currently the starting 11 at Arsenal includes two
describe the institutional setting; in Section III,         English players. Two of the world’s top ten transfer
we survey the literature on the topic; in Section IV,       records belong to Manchester United, and, over the
we present the theoretical framework; in Section V,         past two seasons, Chelsea has spent over £200 million
the data and results are presented; in Section VI, the      on new players. Not only do these big clubs have
efficiency rankings are presented; in Section VII,          to compete in Europe, but also in traditional cup
we discuss the results and, finally, in Section VIII        competitions (such as the League Cup and the FA
we draw conclusions.                                        Cup) and in the domestic league, which tends to
                                                            cause some clubs to field ‘reserve’ sides in the less
                                                            glamorous competitions. Thus, it appears that some
                                                            clubs create alternative squads, with one set of
II. Institutional Setting
                                                            players competing in less prestigious games, whilst
                                                            the ‘superstars’ play in the more celebrated and
The English Premier League is the most profitable
                                                            lucrative matches.
football league, not only in Europe but also world-
                                                               Hence, the finance and performance of these
wide, and contains the world’s richest club:
                                                            leading clubs can be very complex indeed. Some
Manchester United. Prior to 1992, there were four
                                                            clubs have started expanding into new markets and
divisions grouped under one league in England.
However, during the late 1980s and early 1990s, the         entered into sponsorship deals. Manchester United
top teams in England sought to improve their share          has entered into a marketing agreement with the
of television broadcasting revenue and became less          New York Yankees, and Arsenal recently signed
willing to subsidise smaller teams through the              a deal with Emirates Airlines amounting to £100
redistribution of this television income. However,          million for stadium and shirt sponsorship. Football in
even though the English Premier League is the most          England is big business, and vital to the success of
prosperous, it is not uncommon to find genuine              this business is success – not only on the pitch,
concern for individual clubs’ financial health.             but also financial success and the development
Not only does the threat and subsequent effect of           of an efficiently produced product. In Table 1,
relegation to the smaller and less lucrative First          we present 12 English football clubs that
Division sometimes leave former Premiership clubs           remained in the Premiership throughout the seasons
(e.g. Derby County and Bradford City) near financial        analysed.
ruin, but they also run the risk of gambling and               Table 1 shows that Manchester United ranks first
missing out on lucrative European championships             in terms of points, which translates into the team’s
such as the Champions League and UEFA Cup.                  position at the end of the season, followed by Arsenal
Most notably, Leeds United invested heavily in              and Newcastle. Leeds ranks first in the ratio of
playing talent only to miss out on qualification for        wages/points, followed by Manchester. Finally,
the Champions League in 2002, thus eliminating              Liverpool ranks first in turnover, followed by
a large sum of expected income and leading to a             Manchester United.
large-scale sell-off of playing talent. During the             These rankings establish a positive correlation
turmoil, Leeds were ultimately relegated. So, achiev-       between turnover, wages and position, signifying
ing success requires spending, but with that                that sports results and financial results are closely
comes risk and it is not uncommon to see many               related.
Technical efficiency in the EPL                                                                                          733
               Table 1. Figures in 2002/03 season

               Football club            Points      Wages (£m)       Ratio wages/points         Turnover (£m)
               Arsenal                  78          60 569            777                         103 801
               Aston Villa              45          32 310            718                          45 447
               Chelsea                  67          54 365            811                          93 027
               Everton                  59          29 735            504                          46 781
               Leeds United             47          56 595           1204                          64 005
               Liverpool                64          54 431            850                       1 013 981
               Manchester United        83          79 517            958                         174 936
               Middlesbrough            49          29 428            601                          40 229
               Newcastle United         69          45 195            655                          96 689
               Southampton              52          26 666            513                          48 875
               Tottenham Hotspur        50          38 024            760                          66 506
               West Ham United          42          33 342            794                          51 712
               Source: Deloitte & Touche (2004).

III. Literature Survey                                           two articles using the stochastic econometric frontier
                                                                 are, in our view, clearly insufficient for analysing such
There are two contemporary approaches to measur-                 an important issue in the sports market context. With
ing efficiency: first, the econometric or parametric             the present article, we seek to widen the scope of
approach, and, second, the nonparametric or                      sports economics in this specific respect and to draw
DEA approach. Unlike the econometric stochastic                  the attention of other researchers to this neglected
frontier approach, DEA permits the use of multiple               aspect of sports management.
inputs and outputs, but does not impose any
functional form on the data, neither does it make
distributional assumptions for the inefficiency term.
Both methods assume that the production function of              IV. Theoretical Framework
the fully efficient decision-making unit is known. In
practice, this is not the case and the efficient isoquant        In this article, we adopt the stochastic cost econo-
must be estimated from the sample data. Under such               metric frontier approach. The frontier approach, first
conditions, the frontier is relative to the sample               proposed by Farrell (1957), was based on cost
considered in the analysis.                                      functions and came to prominence in the late 1970s
   An important advantage of the econometric                     as a result of the work of Aigner et al. (1977), Battese
frontier is that there are a number of well-developed            and Corra (1977) and Meeusen and van den Broeck
statistical tests available for investigating the validity       (1977). The adequacy of a cost or production
of the model specification – tests of significance for           function depends on the environment in which the
the inclusion or exclusion of factors, or for verifying          units analysed operate. In an environment where
the functional form. The accuracy of these hypotheses            the ultimate objective is to maximise sales and profits,
depends to some extent on the assumption of                      the producers face exogenously determined input
normality of errors, which is not always fulfilled.              prices and output prices and attempt to allocate
A second advantage of the econometric frontier is                inputs and outputs so as to maximise sales. Assuming
that if a variable which is not relevant is included, it         this is the main strategy at football clubs, the
will have a low or even zero weighting in the calcu-             production frontier is the most adequate model for
lation of the efficiency scores, so that its impact is           analysing efficiency (Kumbhakar, 1987). However,
likely to be negligible. This is an important difference         when we have several outputs, it is better to adopt a
from DEA, where the weights for a variable are                   cost frontier approach, relying on the duality theory
usually unconstrained. A third advantage of the                  (Cornes, 1992).
econometric frontier is that it permits the decom-                  The general frontier cost function, which is dual to
position of deviations from efficient levels into ‘noise’        the production function proposed by Aigner et al.
(or stochastic shocks) and pure inefficiency, while              (1977) and Meeusen and Van den Broeck (1977),
DEA classifies the whole deviation as inefficiency.              is as follows:
   Table 2 lists the characteristics of the articles
reviewed.                                                              Costit ¼ 0t þ it Pit þ it Yit þ ðVit þ Uit Þ
   Nine articles using DEA, three articles using
a deterministic econometric frontier approach and                            i ¼ 1, 2, . . . , N; t ¼ 1, 2, . . . , N    ð1Þ
734

Table 2. Literature survey of frontier models on sports
Articles                       Method                     Units                       Inputs                      Outputs                     Prices
Barros and Santos (2005)       DEA-CCR model and          Soccer clubs in the first   Supplies and services       Match receipts, mem-        –
                                DEA-BCC model               Portuguese league           expenditures, wage          bership receipts,
                                                                                        expenditures, amorti-       sponsorship receipts,
                                                                                        zation expenditure,         TV receipts, gains on
                                                                                        other costs                 players, financial
                                                                                                                    receipts, points won,
                                                                                                                    tickets sold
Haas (2003a)                   DEA-CCR and DEA-           12 USA soccer clubs         Players’ wages, coaches’    Points awarded, number      –
                                BCC model                   observed in year 2000       wages, stadium utili-       of spectators and total
                                                                                        zation rate                 revenue
Haas (2003b)                   DEA-CCR and DEA-           20 English Premier          Total wages, coache’s       Points, spectators and      –
                                BCC model                   League clubs                salary, home town           revenue
                                                            observed in 1 year          population
                                                            (2000/01)
Barros and Santos (2003)       DEA-Malmquist index        18 training activities of   Number of trainers,         Number of participants,
                                                            the sports federations,     trainers’ remunera-         number of courses,
                                                            1999–2001                   tion, number of             number of approvals
                                                                                        administrators,
                                                                                        administrators’ remu-
                                                                                        neration and physical
                                                                                        capital
Barros (2003)                  DEA-allocative model       19 training activities of   Number of trainers,         Number of participants,     Price of trainers, price of
                                                            the sports federations,     number of adminis-          number of courses,          administrators, price
                                                            1998–2001                   trators, physical           number of approvals         of capital
                                                                                        capital
Fizel and D’Itri (1997)        DEA-CCR model in           147 college basketball      Player talent, opponents’   Winning percentages         –
                                first stage and regres-     teams, 1984 to 1991         strength
                                sion analysis in
                                second stage
Fizel and D’Itri (1996)        DEA-CCR model              Baseball managers           Player talent, opponents’   Winning percentages         –
                                                                                        strength
                                                                                                                                                                            C. P. Barros and S. Leach
Porter and Scully (1982)    A linear programming      Major league baseball      Team hitting and team         Team percentage wins     –
                              technique (probably      teams, 1961 to 1980         pitching
                              DEA-CCR)
Dawson et al. (2000)        Stochastic Cobb–          Sample of English          Player age, career league     Winning percentages      –
                              Douglas frontier          football managers,         experience, career
                              model                     1992 to 1998               goals, number of pre-
                                                                                   vious teams, league
                                                                                   appearances in the
                                                                                   previous season, goals
                                                                                   scored, player divi-
                                                                                   sional status
Hadley et al. (2000)        Deterministic frontier    National football league   Twenty-four indepen-          Team wins                –
                              model                     teams, 1969/70 to          dent variables
                                                                                                                                            Technical efficiency in the EPL

                                                        1992/93                    describing attack and
                                                                                   defence.
Audas et al. (2000)         Hazard functions          English professional       Match result, league          Duration (measured by
                                                        soccer, 1972/73 to         position and manager          the number of league
                                                        1996/97, match level       age, manager experi-          matches played)
                                                        data                       ence, player
                                                                                   experience
Hoeffler and Payne (1997)   Stochastic production     27 NBA teams, 1992–        Ratio of field goal per-      Actual number of wins
                              frontier                  1993                       centage, ratio of free
                                                                                   throw percentage,
                                                                                   ratio of offensive
                                                                                   rebounds, ratio of
                                                                                   defensive rebounds,
                                                                                   ratio of assists, ratio
                                                                                   of steals, ratio of
                                                                                   turnover and differ-
                                                                                   ence in blocked shots
Scully (1994)               Deterministic and         41 basketball coaches,     Team hitting and team         Winning percentages      –
                              stochastic Cobb–          1949/50 to 1989/90         pitching
                              Douglas frontier
                              model
Zak et al. (1979)           Cobb–Douglas determi-     National basketball        Ten variables of pitch        Ratio of final scores    –
                              nistic frontier model     association teams          performance such as
                                                                                   ratio of steals, ratio of
                                                                                   assists
                                                                                                                                             735
736                                                                                      C. P. Barros and S. Leach
where Cit represents a scalar cost of the i decision-       inefficiency, Zit should be included in the cost
making unit under analysis in the t-th period, Pit is       function.
a vector of input prices, and Yit is a vector of output        The parameters of the model (, ,  and ) are
descriptors used by the i-th club in the t-th period.       estimated using the maximum-likelihood estimator;
   The error term Vit is the traditional error term         the likelihood function can be found in Battese and
of econometric models, assumed to be independently          Coelli (1988). Thus, the technical inefficiency of the
and identically distributed, which represents the effect    i-th club at time t is:
of random shocks (noise) and is independent of Uit.
                                                                      TEit ¼ exðUit Þ ¼ expðzit   Wit :          ð4Þ
   The inefficient term Uit represents technical ineffi-
ciencies and is assumed to be positive and distributed      The conditional expectation of TE is defined under
normally with zero mean and variance U2 . The              the half-normal assumption:
Uit positive disturbance is reflected in a half-normal                                            "           #
independent distribution truncated at zero,                                 Ui                       i =i
                                                                    E                     ¼ i þ  i            ð5Þ
Nðmit , U2 Þ, signifying that each club’s production                 "i1 , . . . , "it               i =i
must lie on or above its cost frontier, but above the
level of one. This implies that the two effects, the V      where ip¼      i  þ ð1 ffi i Þð"i Þ, i ¼ 1=ð1 þ ð=Ti Þ
                                                                         ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
effect, which is a random shock, and the U effect,          and i ¼ U2 =ð1 þ Ti Þ.  is the mean value of the
which is a management shock controlled by the               distribution and T is the time period of the panel,
office, cause any deviation from the frontier.               is the standard normal distribution, and  is the
   The mean inefficiency of the technical efficiency        respective cumulative distribution function (Coelli
effects model, in Coelli et al. (1998) is a deterministic   et al., 1998; Kumbhakar and Lovell, 2000).
function of p explanatory variables:
                       mit ¼ zit                     ð2Þ
where  is a p  1 vector of parameters to be               V. Data
estimated. Following Battese and Corra (1977),
the total variance is defined as  2 ¼ V2 þ U2 . The      To estimate the cost frontier, we used balanced panel
contribution of the error term to the total variation is    data on English Premier League Football clubs in the
as follows: V2 ¼  2 =ð1 þ 2 Þ. The contribution of the   years 1998/99 to 2002/03 (12 clubs 5 years ¼ 60
inefficient term is as follows: U2 ¼  2 2 =ð1 þ 2 Þ,    observations).
where V2 is the variance of the error term V, U2 is the      Frontier models require the identification of inputs
variance of the inefficient term U and  is defined         (resources) and outputs (transformation of
as  ¼ U =V , providing an indication of the relative     resources). Several criteria can be used. Firstly, one
contribution of U and V to ".                               empirical criterion is the availability of data. It is
   The inefficiencies in Uit in Equation 1 can be           important for the applicability of the model results
specified as:                                               that football clubs ‘buy in’ to the process, that the
                    Uit ¼ zit  þ Wit                 ð3Þ   measures of inputs and outputs are relevant, that
                                                            the appropriate archival data are available and that
where Wit is defined by the truncation of the normal        ‘more is better’ in the case of outputs. Usually the
distribution with mean zero and variance 2. Using          criterion of available archival data is used, since it
this parameterisation, a test can be constructed to         encompasses all the previous criteria and therefore
determine whether the estimated frontier is actually        means that the availability of data is the first criterion
stochastic;  ¼ 0 implies that the variance associated      in input and output selection. Secondly, the literature
with the one-sided (efficiency) errors, U2 , is zero,      survey is a way of ensuring the validity of the research
meaning that these deviations from the frontier are         and therefore another criterion to be taken into
better represented as fixed effects in the production       account. The final criterion for measurement selec-
function. Therefore, a test of the null hypothesis that     tion is the professional opinions of sports managers.
 ¼ 0 against the alternative hypothesis that  is          In this article, we follow these three criteria.
positive is used to test whether deviations from the           Based on the available data span, we estimated
frontier are stochastic and whether one should              a generalized Cobb–Douglas stochastic cost function.
proceed with the estimation of parameters related to        We transformed the variables in keeping with
the sources of inefficiency within the context of a         the description column of Table 3. We adopted
stochastic production frontier. Failure to reject the       the traditional log–log specification to allow for the
null hypothesis suggests that the determinants of           possible nonlinearity of the frontier.
Technical efficiency in the EPL                                                                                          737
Table 3. Descriptive statistics of the data

Variable              Description                                         Minimum         Maximum          Mean         SD
Log cost              Logarithm of operational cost in pounds at             4.23         5.16             4.741        0.188
                        constant price 1999 ¼ 100
Log PL                Logarithm of price of players measured by              2.55         3.49             3.053        0.217
                        dividing total wage by the number of players
Log PK1               Logarithm of price of capital measured by              2.08         2.99             2.612        0.202
                        dividing the amortisation of players by the
                        number of players
Log PK2               Logarithm of price capital measured by divid-       0.30           1.89             0.886        0.499
                        ing stadium facilities expenditures by net
                        assets and liabilities
Log points            Logarithm of the points obtained in the season         1.61         1.96             1.760        0.095
Log attendance        Logarithm of the number of tickets sold in the         4.18         4.83             4.556        0.127
                        season
Log turnover          Logarithm of turnover in the season, pounds            4.13         6.01             4.736        0.279
                        at constant price 1999 ¼ 100
Log population        Logarithm of the population in the city of the         5.830        7.016            6.649        6.618
                        club
Log income            Logarithm of the income of the city of the             4.144        4.507            4.342        3.742
                        club, pounds at constant price 1999 ¼ 1000
European              Dummy variable, which is one for clubs                 0            1                0.33         –
                        participating in the European cups in each
                        season

   We noted that the range is narrow, indicating that            Table 4. Stochastic Cobb–Douglas panel cost frontier model
the clubs in the sample are of a similar size in terms
                                                                 Variables                              Coefficients (t-ratio)
of inputs and outputs, but that there is a very wide
difference in the population and income of the club              Constant (0)                           2.198 (2.473)
                                                                 Log PL (1)                             0.872 (3.185)
bases. The rationale for using capital-players needs
                                                                 Log PK1 (2)                            0.101 (2.777)
a justification. Football clubs use players as an active,        Log points (3)                        0.543 (1.840)
tradable commodity in order to capitalise on their               Log attendance (4)                    0.301 (2.085)
market value. Moreover, football clubs are allowed               Log Turnover (5)                       0.400 (3.064)
to amortise the value of the football player on the              Constant (0)                           0.899 (0.190)
                                                                 Log population (1)                    0.308 (1.260)
balance sheet.                                                   Log income (2)                         0.303 (3.553)
                                                                 European (3)                          0.046 (1.172)
                                                                  2 ¼ V2 þ U
                                                                              2
                                                                                                         0.369 (4.939)
Results                                                           ¼ U2 = 2                            0.0006 (3.868)
                                                                 Log(likelihood)                        13.693
In this study, we estimated a Cobb–Douglas stochas-              Lagrange test                           0.366
tic cost function with three input prices (one price             Observations                           60
of labour and two prices of capital), three outputs              Notes: Dependent variable log of total cost. t-Statistics in
(points, attendance and turnover) and contextual                 parentheses are below the parameters, those followed by *
variables (population and income in the club area and            are significant at the 1% level.
whether or not the club is playing in the European
leagues).
                                              
       Cit                   PLit            PK1it               of efficiency that are controlled by management
log          ¼ 0 þ 1 log          þ 2 log                     (labour, capital, attendance and turnover) and for
     PK2it                  PK2it            PK2it
                                                                 the contextual factors that are beyond managerial
                þ 3 logðPointsÞit þ 4 LogðAttendanceÞ
                                                                 control (population, income, European). The vari-
                þ 5 log Turnoverit þ ðVit þ Uit Þ               ables were defined and characterized in Table 3.
                              
Ui ¼ 0 þ 1 log Populationit                                    Table 4 presents the results obtained for the
                                                               stochastic frontier using Frontier 4.1 from Coelli
     þ 2 logðIncomeit Þ þ 3 log Europeanit         ð6Þ
                                                                 (1996), with a half-normal distribution specification.
   This is the cost frontier model, known as the                    We can see that the Cobb–Douglas cost function
technical efficiency effects model, found in Coelli              specified above fits the data well, as the R-squared
et al. (1998), because it accounts for the causes                from the initial ordinary least-squares estimation that
738                                                                                   C. P. Barros and S. Leach
was used to obtain the starting values for the             and attendance. This means that it is costly for
maximum-likelihood estimation is in excess of 85%          football clubs to generate turnover from their
and the overall F-statistic is 282.02. We can also         activity. However, such generation of turnover is
see that the variables have the expected signs, with       independent of performance on the pitch, since
the operating cost increasing with the price of labour     points and attendance contribute negatively to
and the price of capital-players. Moreover, the total      costs. Moreover, the contextual variables play a role
cost decreases with points, attendance, population         in this context, with the population of the club base
and European. Finally, the total cost increases with       contributing negatively to costs, reflecting support
the price of labour, the price of capital, turnover and    for the club both through the contributions of season
income. The frontier parameters are all statistically      ticket holders and through attendance. Finally,
significant and the inefficient error term () is 0.6%     participation in European competitions also contri-
of the total variance, which is a low value when           butes negatively to costs, as a result of the bonuses
compared with other industries, such as banking            earned in European competitions.
(Drake and Weyman-Jones, 1996; Ashton, 2001).                 The income of the club base is, however, positively
                                                           related to costs, reflecting congestion costs in the
                                                           big cities. Secondly, scale in a Cobb–Douglas func-
VI. Efficiency Rankings                                    tion is defined as the sum of the parameters and,
                                                           in the present case, the cost elasticity is equal to
Table 5 presents the results of the time-invariant         0.429 at the sample mean, signifying decreasing
efficiency scores computed from the residuals.             returns to scale. Thus, a 10% increase in outputs
Technical efficiency is achieved, in a broad economic      leads to about a 4.29% increase in costs. The inverse
sense, by the unit which allocates resources without       of this is larger than unity, indicating increasing
waste, and thus the concept refers to a situation          returns to scale on production. This result means that
on the frontier. A unit with a score equal to one is on    scale is a major issue in the football industry, a result
the frontier and those with a score lower than one         confirmed by the DEA research (Haas, 2003b).
are above the cost frontier of best practices. The value   Thirdly, we can see that the elite clubs (Manchester
of waste is measured by the difference between             United, Arsenal and Chelsea) are the least efficient
one and the score, so that, for example, the waste         whenever we include contextual variables in the
of Arsenal is 1  0.901 ¼ 0.099. This is a small waste     analysis. The meaning of this ranking, which includes
when compared with the values observed in other            sports and financial questions, is that despite being
sectors of activity.                                       league champions, these clubs use too many resources
   We can see that the mean score is 96%. This score       to win, and therefore they have a tendency to perform
suggests that football clubs could reduce their output     very well in terms of sport but not in terms of finance.
cost by 3% without decreasing their input, which,          The second-tier clubs, those clubs which, despite not
in this case, is the price of labour and the price         winning, represent the country in the European cups
of capital-players. The maximum football club              (e.g. Liverpool and Newcastle), are better positioned
score was naturally 1, which was achieved by               in the efficiency rankings. However, the best situation
Middlesbrough, while the minimum efficiency score          is obtained by the third-tier clubs, those clubs which
was 96% and was achieved by Arsenal in the first           are playing in sub-championships of their own, with
three years and then by Chelsea. The median was            very different objectives from the few elite clubs.
97%, which was smaller than the mean. Therefore,           Clubs such as Middlesbrough and Southampton
there are more clubs below the mean than above             manage their position in the league prudently.
the mean. The SD was 1.3%. These efficiency scores         Therefore, the general conclusion is that the three
are high in comparison with those found elsewhere          clusters of clubs observed have different managerial
in other activities, such as banking (Ashton, 2001).       objectives, and that scale, overspending and manage-
High efficiency scores are consistent with efficient       rial skills are necessary to win the league. Fourthly,
and more competitive organisations, such as those          the mean efficiency of the clubs in the league is
observed in sports.                                        relatively high, when compared with other industries
                                                           (Ashton, 2001). This signifies that, on the pitch,
                                                           football clubs are scrutinized to the extreme. In the
VII. Discussion                                            stock exchange, football clubs are scrutinised along-
                                                           side the other quoted firms. Therefore, despite some
What is the meaning of these results? Firstly, it          failures observed, this is an industry that is much
can be seen that the cost increases with all factors       more closely scrutinised than the average organisa-
of production, with the exception of points                tion, which results in a high level of efficiency.
Table 5. Efficiency scores
                                                                                                                                                                         Technical efficiency in the EPL

                       Efficiency scores                          Efficiency scores                          Efficiency scores                       Efficiency scores
Football clubs         in 1998/99          Football clubs         in 1999/2000        Football clubs         2000/01             Football clubs      2002/03
Middlesbrough          1.0000              Middlesbrough          1.0000              Middlesbrough          1.0000              Middlesbrough       1.0000
Southampton            0.9924              Newcastle              0.9939              Southampton            0.9932              Southampton         0.9946
Newcastle              0.9924              Southampton            0.9927              Newcastle              0.9930              Leeds               0.9940
Liverpool              0.9911              Everton                0.9914              Everton                0.9910              Newcastle           0.9936
Everton                0.9911              Aston Villa            0.9876              Aston Villa            0.9879              Everton             0.9924
Aston Villa            0.9841              Liverpool              0.9750              Tottenham              0.9731              Aston Villa         0.9890
Tottenham              0.9726              Chelsea                0.9732              West Ham               0.9731              Tottenham           0.9741
West Ham               0.9726              West Ham               0.9731              Chelsea                0.9731              West Ham            0.9741
Leeds                  0.9659              Tottenham              0.9730              Leeds                  0.9648              Liverpool           0.9653
Manchester United      0.9657              Leeds                  0.9653              Liverpool              0.9648              Manchester United   0.9651
Chelsea                0.9653              Manchester United      0.9652              Manchester United      0.9646              Arsenal             0.9648
Arsenal                0.9653              Arsenal                0.9648              Arsenal                0.9643              Chelsea             0.9648
Mean                   0.9799              –                      0.9796              –                      0.9786              –                   0.9810
Median                 0.9784              –                      0.9741              –                      0.9731              –                   0.9815
SD                     0.0131              –                      0.0126              –                      0.0134              –                   0.0141
Note: The efficiency scores for the 2001/02 season are not displayed, but are available under request from the authors.
                                                                                                                                                                          739
740                                                                                     C. P. Barros and S. Leach
The emotional discourse that surrounds the game              A policy for overcoming the identified inefficiencies
clouds the efficiency drive the industry has adopted.      should start with an analysis of the scale of activities
   How do we explain the different strategies adopted      and the adoption of a competitive sporting strategy,
by the identified cluster of football clubs? These         as the case of Leeds has shown in confirming the
different strategies stem from strategic-based groups      theoretical results, in which the population of the club
and differences in terms of resources. Firstly,            base is a main driver in economic performance
strategic-based groups (Caves and Porter, 1977)            (El-Hodiri and Quirk, 1971; Fort and Quirk, 1995).
refer to differences in the structural characteristics
of units within an industry, which lead to differences
in performance. In football, clubs with similar asset
configurations pursue similar strategies with similar      VIII. Conclusion
results in terms of performance (Porter, 1979). While
there are different strategic options among the sectors    This article has proposed a simple framework for the
of an industry, because of mobility impediments, not       evaluation of English Football Premier League Clubs
all options are available to each industry, causing        and the rationalisation of their operational activities.
a spread of the efficiency scores in the industry.         The analysis is based on a stochastic frontier model.
Secondly, the differences in resources available to        Benchmarks are provided for improving the opera-
clubs (Wernerfelt, 1984; Barney, 1991; Rumelt, 1991)       tions of sub-optimal performing clubs. Several inter-
mean that football clubs are heterogeneous in relation     esting and useful economic insights and implications
to the resources and capabilities on which they base       are raised by the study. For the group with the lowest
their strategies. These resources and capabilities may     efficiency score, adjustment is needed in order to
not be perfectly mobile across the industry, resulting     reach the efficiency frontier. Too much expenditure
in a competitive advantage for the best-performing         on factors adds to inefficiency, namely when such
football clubs.                                            expenditure is not translated into points. Attendance
   Purchasable assets cannot represent sources of          and turnover increases translate into cost increases,
sustainable profits. Indeed, critical resources are not    so that managerial procedures to decrease the
available on the market. Instead, they are built up        contribution of these outputs to costs might be a
and accumulated on the football club’s premises,           priority for English football managers.
                                                              The general conclusion is that football clubs have
their nonimitability and nonsubstitutability being
                                                           different efficiency scores. Sporting success is a main
dependent on the specific traits of their accumulation
                                                           driver in cost control, together with scale, confirming
process. The difference in resources thus results in
                                                           the importance of the local fan base. Managerial
barriers to imitation (Rumelt, 1991) and in the
                                                           skills are also important and explain part of the
football clubs’ inability to alter their accumulated
                                                           behaviour observed. The role played by managerial
stock of resources over time. In this context, unique
                                                           skills in sports is linked to sporting and financial
assets are seen as exhibiting inherently differentiated
                                                           performance in the football market.
levels of efficiency; sustainable profits are ultimately
                                                              More investigation is needed to address the
a return on the unique assets owned and controlled
                                                           limitations mentioned.
by the club (Teece et al., 1997).
   Accordingly, football clubs may achieve high levels
of competitiveness by using vast amounts of resources
and thus perform inefficiently, or they may achieve        References
low levels of competitiveness and perform efficiently.     Ashton, J. K. (2001) Cost efficiency characteristics of
   The general conclusion is that sporting success is          British retail banks, The Service Industries Journal, 21,
a main driver in cost control, together with scale,            159–74.
confirming the importance of the local fan base.           Aigner, D. J., Lovell, C. A. K. and Schmidt, P. (1977)
Managerial skills are also important and explain               Formulation and estimation of stochastic frontier
                                                               production function models, Journal of Econometrics,
part of the behaviour observed. The role played by             6, 21–37.
managerial skills in sports is linked to sporting and      Audas, R., Dobson, S. and Goddard, J. (2000)
financial performance in the football market, decreas-         Organizational performance and managerial turnover,
ing costs and increasing sporting performance in               Managerial and Decision Economics, 20, 305–18.
order to perform better on the financial side. The role    Barney, J. (1991) Firm resources and sustained competitive
                                                               advantage, Journal of Management, 17, 99–120.
played by the club’s social base in terms of popula-       Barros, C. P. (2003) Incentive regulation and efficiency
tion, income and scale should also be taken into               in sports organisational training activities, Sport
account.                                                       Management Review, 6, 33–52.
Technical efficiency in the EPL                                                                                         741
Barros, C. P. and Santos, A. (2003) Productivity in sports      Fizel, J. L. and D’Itri, M. P. (1997) Managerial efficiency,
     organisational training activities: a DEA study,                managerial succession and organizational perfor-
     European Journal of Sport Management Quarterly, 1,              mance, Managerial and Decision Economics, 18,
     46–65.                                                          295–308.
Barros, C. P. and Santos, A. (2005) Les relations entre la      Fort, R. and Quirk, J. (1995) Cross-subsidization, incen-
     performance sportive et la performance financière:             tives, and outcomes in professional team sports
     application au cas du football Porugais, in Marketing           leagues, Journal of Economic Literature, 33, 1265–99.
     and Football: Une Perspective International (Eds)          Haas, D. J. (2003a) Technical efficiency in the major league
     G. Bolle and M. Desbordes, Presses Universitaires du            soccer, Journal of Sport Economics, 4, 203–15.
     Sport, Voiron, France, pp. 347–68.                         Haas, D. J. (2003b) Productive efficiency of english football
Battese, G. E. and Coelli, T. J. (1988) Prediction of firm-          teams – a data envelopment approach, Managerial and
     level technical efficiencies with a generalised frontier        Decision Economics, 24, 403–10.
     production function and panel data, Journal of             Hadley, L., Poitras, M., Ruggiero, J. and Knowles, S.
     Econometrics, 38, 387–99.                                       (2000) Performance evaluation of National football
Battese, G. E. and Corra, G. S. (1977) Estimation of                 league teams, Managerial and Decision Economics, 21,
     a production frontier model: with application to the            63–70.
     pastoral zone of Eastern Australia, Australian Journal     Hoeffler, R. A. and Payne, J. E. (1997) Measuring
     of Agricultural Economics, 21, 169–79.                          efficiency in the National basketball association,
Caves, R. and Porter, M. E. (1977) From entry barriers               Economic Letters, 55, 293–99.
     to mobility barriers: conjectural decisions and con-       Kumbhakar, S. C. (1987) Production frontiers and panel
     trived deterrence to new competition, Quarterly                 data: an application to US class 1 railroads, Journal of
     Journal of Economics, 91, 241–61.                               Business and Economics Statistics, 5, 249–55.
Cornes, R. (1992) Duality and Modern Economics,                 Kumbhakar, S. C. and Lovell, C. A. K. (2000) Stochastic
     Cambridge University Press, Cambridge.                          Frontier Analysis, Cambridge University Press,
Coelli, T. J. (1996) A Guide to FRONTIER Version 4.1:                Cambridge, UK.
     A Computer Program. For Stochastic Frontier                Meeusen, W. and van den Broeck, J. (1977) Efficiency
     Production and Cost Function estimation. Working                estimation from a Cobb–Douglas production function
     Paper No. 7/96, Centre for Efficiency and Productivity
                                                                     with composed error, International Economic Review,
     Analysis. University of New England, Armidale,
                                                                     18, 435–44.
     Australia.
                                                                Porter, M. E. (1979) The structure within industries and
Coelli, T. J., Rao, P. and Battese, G. E. (1998) An
                                                                     companies’ performance, The Review of Economics and
     Introduction to Efficiency and Productivity Analysis,
     Kluwer Academic Press, Boston.                                  Statistics, 61, 214–27.
Dawson, P., Dobson, S. and Gerrard, B. (2000) Stochastic        Porter, P. and Scully, G. W. (1982) Measuring managerial
     frontier and the temporal structure of managerial               efficiency: the case of baseball, Southern Economic
     efficiency in English soccer, Journal of Sports                 Journal, 48, 642–50.
     Economics, 1, 341–62.                                      Rumelt, R. (1991) How much does industry matter?,
Drake, L. and Weyman-Jones, T. G. (1992) Technical and               Strategic Management Journal, 12, 167–85.
     scale efficiency in UK building societies, Applied         Scully, G. W. (1994) Managerial efficiency and survivability
     Financial Economics, 2, 1–9.                                    in professional team sports, Managerial and Decision
El-Hodiri, M. and Quirk, J. (1971) An economic model                 Economics, 15, 403–11.
     of a professional sports league, Journal of Political      Teece, D., Pisano, G. and Shuen, A. (1997) Dynamic
     Economy, 79, 1302–19.                                           capabilities and strategic management, Strategic
Farrell, M. J. (1957) The measurement of productive                  Management Journal, 18, 509–33.
     efficiency, Journal of the Royal Statistical Society,      Wernerfelt, B. (1984) A resource-based view of the firm,
     Series A, 120, 253–90.                                          Strategic Management Journal, 5, 171–80.
Fizel, J. L. and D’Itri, M. P. (1996) Estimating managerial     Zak, T. A., Huang, C. J. and Siegfried, J. J. (1979)
     efficiency: the case of college basketball coaches,             Production efficiency: the case of professional basket-
     Journal of Sport Management, 10, 435–45.                        ball, Journal of Business, 52, 379–92.
You can also read