Buying Success? The Marginal Effect of Transfer Spending on the performance of Premier League clubs

  • Upload
    hugh

  • View
    7

  • Download
    0

Embed Size (px)

DESCRIPTION

Dissertation submitted in partial requirements for the Bachelor of Arts Degree, School of Economics, University College Dublin, April 2010

Citation preview

Research Project Draft

Buying Success?

The Marginal Effect of Transfer Spending on the performance of Premier League clubs

Hugh McDowell06580360

Dissertation submitted in partial requirements for the Bachelor of Arts Degree, School of Economics, University College Dublin, April 2010

TABLE OF CONTENTSPage

TITLE PAGEi

TABLE OF CONTENTSii

ABSTRACTiii

1: INTRODUCTION1

1.1Purpose and motivations1

1.2Contextualization the Premier League and the transfer window2

1.3Contribution of the research2

1.4Organization of the paper3

2: LITERATURE REVIEW3

2.1Buying success in the Premier League3

2.2Human capital in sport4

3: RESEARCH QUESTIONS AND RESEARCH METHODOLOGY5

3.1Issue 1: The marginal effect of spending on performance5

3.2Issue 2: The marginal effect of money spent in each transfer window5

3.3Issue 3: The lagged effect of spending on performance5

3.4Population and data collection6

3.5Research methodology6

3.5.1Measurement of independent variable6

3.5.2Measurement of dependent variable7

3.5.3Predicted response of dependent variable to independent variable8

3.5.4Method9

4: RESULTS10

4.1The marginal effect of total net spend on points11

4.2The marginal effects of Summer and January net spend12

4.3The lagged effect of net spending12

5: SUMMARY AND IMPLICATIONS OF RESULTS13

5.1Objectives13

5.2Discussion and analysis of results14

5.3Limitations of research18

5.4Suggestions for further research20

5.5Conclusion21

REFERENCES 22

ABSTRACTPrior research in the production function of a football club has tended to focus either on the club as a financial entity, or on the relationship between payroll size and team performance. This study seeks to focuses on the empirical relationship between the amount of money football teams spend in the transfer market and their subsequent performance in the league. The research is based on an econometric analysis of league points accumulated and outlay in the transfer market.

The paper examines the performance Englands Premier League clubs over a five year period, from 2004 to 2009, with reference to their activity in the transfer market over the same period. Three issues are investigated: the relationship between net transfer spending and Premier League performance over a season; the marginal effects of money spent in the January transfer window and the Summer transfer window; and the lagged effect of transfer activity in a season on performance in subsequent seasons.Findings of the research indicate a positive relationship between spending and performance, and, in particular, a strong marginal effect for summer spending in comparison with spending in January. Furthermore, analysis suggests that investment in playing personnel yields improved team performance both in the season in question and in subsequent seasons.

1: INTRODUCTIONThis project investigates the marginal effect of money spent in the transfer market on the league performance of football teams. During the course of the paper, three issues will be addressed.The first concerns the overall effect of a football clubs net transfer spending during the course of a season on that clubs performance during that season. Team performance is measured by league points accumulated. Secondly, the difference between the marginal effect of spending during the Summer transfer window and marginal effect of spending during the January transfer window is examined. Finally, the lagged effect of net transfer spending is analysed.1.1 Purpose and motivations

The exponential growth in the global interest in world football since the formation of the Premier League in 1992 has made it one of the most popular sporting competitions in the world. The rapid spread of television in developing countries over the last two decades has brought about a remarkable increase the Premier Leagues audience. The competitions increased popularity has injected huge sums of money into English football, and football clubs are now able to generate huge revenues through sponsorship and marketing. The primary purpose of this paper is to investigate to what extent clubs can use their vast earnings, through the acquisition of human capital, to influence their on-field performance.There are several factors which are thought to influence the success of a football team. Some are considered absolute truisms of the sport the quality of coaching, the quality of playing personnel, or the quality of tactics. Others are less tangible the number of home-grown players in a squad, the mentality surrounding the club, the clubs history, or even the size of the clubs stadium. In the past, research papers of this nature have attempted to formalize the rather informal notions about football production functions intuitively (or unconsciously) held by managers. (Carmichael, Thomas & Ward, 2000, p.11). A fiercely contested concept among followers of professional football is the notion that money buys success; that a teams performance will improve as a result of large amounts of spending in the transfer market. Advocates claim that, in the modern era of the game, money has become the sine qua non of football success; critics argue that putting eleven brilliant footballers onto a pitch together does not necessarily produce a good football team, and that a certain level of cohesion and understanding between the players is required. The motivation of this paper is establish whether or not money truly can buy success.1.2 Contextualization the Premier League and the transfer windowThe Premier League is the highest division of English footballs tiered system. Each of its twenty teams plays each other twice in a season, once at home and once away. Teams are awarded three points for a win, one point for a draw and no points for a loss. At the end of the season, which lasts from August until May, the three teams with the least points are relegated to the English Championship, Englands second division, and are replaced with three promoted teams from that division. The team which finishes the season with the most points is crowned champions, which the top four clubs gain qualification to the lucrative European Champions League.In response to negotiations with the European Commission over employment conditions, FIFA, world footballs governing body, imposed a transfer window system in the 2002-2003 season. Only during transfer windows are clubs allowed to buy and sell playing personnel from other clubs. For most major European leagues, including the English Premier League, the transfer window system comprises of a shorter window, generally from January 1st until January 31st (henceforth referred to as the January window), and a longer window, from July 1st until August 31st (henceforth referred to as the Summer window). 1.3 Contribution of the research

This research can be distinguished from existing literature in the field of sport economics in two ways. Firstly, research up until now has established linkages between investment, in the form of payroll, and performance. Footballs transfer market sets it apart from most other team sports in terms of the mechanism by which players move from one team, club or franchise to another. In practically all team sports (and, indeed, professions), transfer fees do not exist, and in ones where they do (rugby union, for example), they are tiny in comparison to those in professional football. On that basis, researchers have tended to examine salaries as a means of investment in personnel. Secondly, the emphasis on the performance of Premier League clubs in previous research has previously tended to pertain to areas such as financial performance and efficiency, rather than being strictly measured by sporting performance. In that regard, this paper is one of few which focus solely on league performance as a measure of success.1.4 Organization of the paper

Section 1 has introduced the aims and context of this research project. Section 2 will examine previous literature in the area of investment in human capital and its effect on performance, with particular emphasis on sport economics. Section 3 will outline the research issues and methodology adopted, and Section 4 will present the results of the research. The paper concludes in Section 5 by considering the implications of the research and the significance of the findings. Limitations of the study and suggestions for further research are also considered.

2: LITERATURE REVIEWLiterature in the area of the economics of team has tended to focus on American sports (in particular, baseball) in which huge amounts of statistical data are readily available. This focus on statistics does not travel across the Atlantic, and on-field performance in European team sports (with the possible exception of cricket) is far more difficult to measure. As a consequence, sports such as football, where performance is almost entirely an output of the collective rather than the individual, have largely been neglected by sport economists. Nevertheless, several studies related to football have produced results relevant to this study.2.1 Buying success in the Premier League

A common research topic in sport economics is to construct production functions for team sports, where in-game statistics are used as input measures. Two such studies which are of relevance to this paper concern Premier League football (Carmichael, Thomas & Ward, 2000) and Rugby League (Carmichael & Thomas, 1995). In the case of the latter, the authors observe that the interdependent nature of team sports make it difficult to identify marginal products for individual players. In both papers, the input co-efficients are highly significant, leading the authors to justify, and highlight the importance of, constructing production functions for sports teams.Several studies of the performance of Premier League clubs suggest that money is influential in success. Leech & Barros (2006) conduct a comparison of each clubs average player wage and its performance which suggests a positive correlation between revenues, wages, and position, signifying that sport results and financial results are related (p.6). Their paper also notes that money cannot be used as an entirely accurate predictor of performance, as the role played by managerial skills in sports is linked to matching sporting and financial performance in the football market.

A contrast of spending and performance (Hall, Szymanski and Zimbalist, 2002) suggests a much closer correlation between the two in football than in baseball. Although Major League Baseball clubs and Premier League clubs had almost identical standard deviations in terms of payroll, the variance in winning percentages for football clubs was over 50% higher despite there being roughly four times more fixtures in Major League Baseball, which would imply that less of its variation is owing to chance. While the Hall et al. measure financial outlay in terms of payroll rather than transfer spending (so that a comparison with baseball can be drawn), they establish a clear Granger causality (which determines whether one time series is useful in forecasting another) between spending and performance. It also goes on to hypothesize that it is the more mobile nature of the footballs player market that makes spending in football more efficient than spending in baseball.

2.2 Human capital in sport

In the absence of a transfer market akin to that found in professional football, economists have used various other statistics to measure the quality of human capital available to a team. Langelett (2003) regresses the performance of college-level American football teams against the quality of player they draft from high school. His research is similar to that in this paper, in that he includes the lagged effect of player recruitment on performance. In this case, human capital is measured using rankings presented by two mainstream media sources after the drafting of high school graduates has taken place. His research demonstrates a clear and significant correlation between the rankings of new recruits and subsequent performance. However, the regression suffers from multi-collinearity between lagged variables. In addition, the issue of causality remains unclear, as there is evidence to suggest that a teams past performance influences their ability to recruit the best players.

3: RESEARCH QUESTIONS AND METHODOLOGYThis section outlines the three issues examined by the research, and describes the manner in which the data is selected, measured and used to address these questions.3.1 Issue 1: The marginal effect of spending on performance

The most obvious, and most important, objective when a football clubs spends money in the transfer market is to improve the quality of its playing personnel. If no correlation between spending and performance existed, it would make no sense for transfer fees to exist. It is intended that the data will establish whether there is a significant marginal effect of spending on performance, whether this effect is positive or negative, and whether it is large enough to justify an aggressive transfer market policy.

3.2 Issue 2: The marginal effect of money spent in each transfer windowWhile the summer window traditionally gives managers a chance to prepare a squad capable of meeting the objectives of the club over the forthcoming season, the January window acts as an opportunity to chop and change based on results achieved in the first half of the season. Having established the marginal effect (if any) of spending on performance, the next issue to be addressed is whether a difference exists in the effectiveness of money spent in the summer transfer window and in the January transfer window.3.3 Issue 3: The lagged effect of spending on performance

The final result which this paper intends to produce is to establish whether transfer spending contains any sort of lagged effect in other words, whether spending in previous seasons will influence outcomes in future seasons. While it is nave to assume that spending during a season ceases to be relevant as soon as that season ends, it is interesting to investigate to what degree time can influence the effectiveness of money spent.3.4 Population and data collection

Due to a growing trend of private ownership of Premier League clubs over the last two decades, the availability of accurate data on transfer fees is very limited. Very few clubs publish the precise details of transfers, and deals often involve a player plus cash trade arrangement, staggered payment schedules, or complex clauses related to, among others, the success of the player, the success of the buying club, or the value of any future sale of the player in question.

In lieu of official data from clubs or the Premier League, this paper will base its calculations on transfer fee estimates from the Soccer Base, a football statistics organization. Soccer Base, formerly known as Stamps Soccer Database, is now owned by betting newspaper The Racing Post, and supplies football statistics and results to British newspapers, television and radio. Soccer Base provides transfer details for all clubs in Great Britain over the last 13 years. While in most cases the sums involved have not been officially verified by the clubs in question or the Premier League, they represent the figures reported in the media and widely believed to be true in the football community.This study uses panel data over five seasons of the Premier League. With twenty teams competing each season, the study contains 100 observations. The data sample for this project consists of the 29 teams who competed in the Premier League from the season 2004-2005 to the season 2008-2009 inclusive. Of these 29 clubs, 14 competed in the Premier League in all of the five seasons. The remaining 15 teams competed in between one and four of the five seasons.

3.5 Research methodologyThis section illustrates the measurements and methods utilised in the research process. It also demonstrates the theoretical model of the production function for each of the three regressions.

3.5.1 Measurement of independent variable

The independent variable for this project is net spend in the player transfer market. In calculating the net spend for a given transfer window, transfer receipts are deducted from transfer payments, with amounts in pounds sterling. This provides us with a net spend figure for each club for each transfer window. Transfer data for the 15 clubs who were not in the Premier League for all five seasons is only included for the season in which they did compete. In the case of any deal which was agreed in principle (and, in some cases, paid for) outside a transfer window, the transfer fee involved is included in the calculations for the window in which the player officially joins the purchasing club.The justification behind the use of net spend, rather than merely expenditure, is that is is assumed that net spend will provide an indicator to a clubs improvement. A club could spend 30 million in the transfer market, but still may not improve if its purchases are financed by sales of players worth 40 million. If it is assumed that, broadly speaking, all clubs value all players the same, it can also be assumed that a net spend of zero will neither improve nor worsen the absolute quality of a clubs squad. However, there are two caveats to this statement. The first is that the assumption of all clubs valuing all players the same is unrealistic. Valuing a football player is a subjective process which is different for each different manager. Furthermore, clubs have varying levels of resources, and the perceived resources of the two clubs involved in a transfer deal will affect the behaviour of both. On the one hand, a so-called mega-rich club, such as Chelsea, is likely face inflated transfer fees for any player it attempts to purchase. On the other hand, a financially-stricken club which desperately needs to offload players to stay afloat will probably receive reduced transfer fees, as rival clubs prey on their eagerness to sell quickly. The second caveat is in a league contest system, the absolute quality of a football squad is not relevant. Because teams score points against each other, an absolute improvement in the quality of a squad will not necessarily earn a team more points if all the other teams in the league improved by a greater degree.3.5.2 Measurement of dependent variable

The dependent variable for this paper is the number of points attained in a Premier League season. This data was also found using Soccer Base. Each of the one hundred observations of points attained corresponds to specific net spend data from the preceding Summer and January transfer windows. In particular, each observation contains a team (designated by a numeric value from 1 to 29), a year, a total net spend, a January window net spend, and a Summer window net spend (where January net spend + Summer net spend = Total net spend). For example, the first observation from the data set:Table 3.1: Sample Data Entry

TeamYear (Season ending May 200X)PointsTotal Net SpendSummer Net SpendJanuary Net Spend

12005831.150.151

Source: www.soccerbase.com

The data entry in Table 3.1 corresponds to team no.1 (Arsenal FC), in the season 2004/2005, during which Arsenal attained 83 points and finished in second place. The club spent net 150,000 in the Summer window 2004, and net 1 million in the January window 2005, leading to a total net spend for the season 2004/2005 of 1.15 million.

3.5.3 Predicted response of dependent variable to independent variable

Modern conventions within football would suggest that a positive relationship between points attained and total net spend should be observed. Football clubs spend money in the transfer market in order to improve the quality and/or quantity of their playing squads. If there was no relationship between expenditure and performance, it would make no sense for clubs to engage in transfer activity. This positive relationship should also be observed in both the January and summer transfer windows. Furthermore, the period of observation coincides with notable improvements in performance on historical trends for several clubs (for example, Chelsea and Aston Villa), which are generally seen as the result of exogenous shocks (Leech & Barros, 2006) - takeovers by wealthy individuals who have subsequently spent heavily in the transfer market. With respect to the differences between January spending and summer spending, there is no prevailing theory which states that one is more effective than the other. One observation worth making at this point is that clubs who have under-performed from August to December might be expected to engage more aggressively in January transfers in an attempt to turn their floundering season around. With respect to the lagged effects of transfer expenditure, conventional wisdom would suggest that it may take more than one season for playing personnel to settle in at a new club (or, in the case of imports, a new league), learn how to play under a new manager and with new players, and adapt to new surroundings. Therefore, it might take a season or more before the full effects of a transfer are borne out in terms of league performance. Similarly, the purchase of young players may display considerable lagged effects, as it might take time for them to break into the team.3.5.4 MethodBefore running regressions, the data is formatted into time series panel data. This combines the cross-sectional examination of the 29 different teams with the chronological element of the five seasons in question.In addressing the first research issue the effect of marginal effect on points of net spending a simple Ordinary Least Squares (OLS) regression of points on total net spend is run. The estimated model is:

,

where is points, c is a constant, x is total net spend, and is an error term. This study focuses on the sign, size and significance of , the co-efficient on total net spend which tells us the marginal effect of spending on points. Its sign signifies whether spending has a positive or negative effect on points, and its size will signify to what degree spending affects points. A t-value of greater than |1.96| implies that the co-efficient is significant at 95%.

The second regression undertaken by this paper will examine the difference between January net spend and summer net spend. The estimated model is similar to the previous model:

where, again, is points, c is a constant and is an error term. In this case, J represents January net spend and S represents summer net spend, with and their respective co-efficients. Similar to the first regression, the sign, size and significance of and signifies the marginal effects of January and summer net spend respectively. The interpretations and significance thresholds remain the same as before.

The third and final regression of this project is slightly different from the previous two, in that it estimates lagged co-efficients. Due to the relatively small time-span of the data, a lag of just two periods is estimated. Any more periods would drop too many data entries to be worthy of analysis. The regression is:

where, as before, is points, c is a constant and is an error term. In addition, is a co-efficient on x, as before, is total net spend. The criteria regarding sign, size and significance remain the same as before. The co-efficients on xn-1 and xn-2 will provide an estimate of the effect on points of total net spend from last season and the season before last, respectively.4: RESULTSThis section details the results of the econometric analysis undertaken in this paper. The section has three parts one for each of the research questions addressed. Table 4.1 contains a brief summary of the data set.Table 4.1: Description of Data

VariableMean MedianStandard deviation

Points52.2849.517.82

Net spend (m)9.83m7.05m16.03m

January net spend (m)2.74m0.38m8.00m

Summer net spend (m)7.08m4.46m13.63m

Number of Observations: 100

4.1 The marginal effect of total net spend on points

Table 4.2: Scatter plot of Total Net Spend on Points

The regression finds a significant and positive marginal effect of net spend. Specifically, our model estimates:

Points = 50.09 + 0.222(total net spend)

In other words, for every million pounds of net spend in a season, a team can expect to attain an extra 0.222 points. With a t-value of 2.02, the co-efficient on total net spend is just about significant at 95%. The standard error on the co-efficient, which is the standard deviation of the difference between the models estimate of the co-efficient and the co-efficients true value, is 0.11.4.2 The marginal effects of summer and January net spend

This regression suggests significantly positive marginal effect for net spending in the summer window, but the co-efficient on January spending is not significant. With 100 observations, the model estimates:Points = 50.12 + .083(January net spend) + .271(summer net spend)

The t-value for the co-efficient on January spending is 0.38, suggesting insignificance to the extent that the result can be ignored. However, the t-value for summer spending is 2.10, which is significant at 95%. It indicates that every 1 million in summer net spending will secure an additional 0.271 points. It is also observed that the standard error on January spending (.220) is noticeably larger than the standard error on summer spending (.129). 4.3 The lagged effect of net spending

In estimating the lag effects of transfer spending, the sample size is 47 and the following is observed:Points = 52.12 0.28(Net Spend season n) 0.81(Net Spend season (n-1)) + .611(Net spend season (n-2))

The inclusion of lags in the model reduces the sample size from 100 to 47, and increases the intercept from just over 50 to 52.12. The reduced sample contains the 14 teams who competed in the Premier League in the five seasons in question, in addition to Charlton Athletic and Wigan Athletic, who each competed for three consecutive seasons, and West Ham United, who competed for four consecutive seasons. Of these 47 observations, only one was relegated (Charlton Athletic in 2007), which explains why the intercept on points is higher than for the previous sample of 100 teams.

The negative co-efficients on this period and last periods total net spend are ignored, as the t-values associated with them (-0.21 and -0.37 respectively) indicate that they are not statistically significant. However, the co-efficient 3, which suggests that a net spend of 1 million two periods ago will gain a team .611 additional points, has a t-value of 3.58 and so is highly significant at 95%.Finally, it is observed that the adjusted R-squared statistic for the final regression is 0.2043, which is far greater than those from either the first (0.0302) or the second (0.0255).Table 4.3: Summary of Regressions

Dependent VariableCo-efficientSignificance Level

Total Net Spend

0.2222.02*

Intercept50.09324.30**

January Net Spend0.2712.10*

Summer Net Spend.0830.38

Intercept50.12324.25**

Lagged Regression

Total Net Spend-.028-0.21

Total Net Spend (n-1)-.081-0.39

Total Net Spend (n-2)0.6113.58**

Intercept52.12316.37**

*Significant at 5%** Significant at 1%5: SUMMARY AND IMPLICATIONS OF RESULTSThis section of the paper summarizes and interprets the results from Section 4, highlights the limitations of the research and suggests topics for future research.

5.1 Objectives

The objective of this research was to establish the effectiveness of transfer market activity for football clubs in terms of on-field performance. In that context, three research questions were addressed:

RQ1: What is the marginal effect of net spend in the transfer market on league performance?RQ2: What are the respective marginal effects of net spend in the January transfer window and in the summer transfer window?RQ3: What is the lagged effect of net spend in previous seasons on league performance?On the basis that significant results were obtained in each of the three research areas, the objectives of this project were met. While the research is not without its limitations (which are addressed later in this section), it also contains worthwhile results which are worthy of analysis.

5.2 Analysis and discussion of results

In accordance with previous literature in this area discussed in the review of literature, a significantly positive marginal effect of expenditure on performance was found. This result confirms the usefulness of the transfer market as a mechanism by which to improve a playing squad. Furthermore, the findings confirm money as a driving factor in the sporting success of a team. In a departure from Hall, Szymanski & Zimbalist (2002), the area of player incentives is not problematic. The magnitude of investment in human capital in the case of this paper should have little bearing on the behaviour of the player, because the size of a transfer fee and a players income are almost completely unrelated.

The magnitude of the marginal effect presents several interesting implications. Firstly, it is sufficiently large to justify large net spends in the transfer market. A net spend of 20 million, for example, will garner roughly 5 additional points an amount sufficiently large to make a difference between qualifying for lucrative European competition and not, or between staying in the Premier League and not. On that basis, it is also worth remarking upon the ability of investment in playing personnel to be a profitable exercise. Indeed, if a considerable outlay during a season thrusts a team into a Champions League qualifying position, it is highly likely that it will have been a profitable venture.

Secondly, the size of the marginal effect re-affirms that money alone cannot buy success. In fact, no team in this study truly buys success; teams contesting the title and teams engaged in relegation fights are both statistical outliers. For example, consider the three lowest and three highest net spends in this data set as shown in Table 5.1.Table 5.1 illustrates a varied cross-section of statistical outliers. Despite relatively large negative net spends, Arsenal (in 2008) finished in 3rd place, while West Ham (in 2009) and Everton (in 2005) were not in danger of relegation. Conversely, while Chelsea (2005) spent a very large amount and won the league, similarly large spends from Tottenham and Manchester City (both 2009) yielded very average points tallies. Even in the case of Chelseas 2005 league title, their net spend under the estimated model would imply of points tally of only 66.9 points, in contrast with their (then) Premier League record of 95 points.

Table 5.1 Biggest and smallest net spends, 2004-2009.

TeamYearPointsTotal Net Spend

Panel A: Three lowest net spends

West Ham200951- 20.8m

Arsenal200883- 20.4m

Everton200561- 16.65m

Panel B: Three highest net spends

Tottenham20095143.05m

Chelsea20059575.85m

Manchester City20095082.4m

As shown in Table 5.2, similar results are observed in the case of the three highest and three lowest points tallies: Table 5.2: Biggest and Smallest Points Tallies, 2004-2009

TeamYearPointsTotal Net Spend

Panel A: Three lowest point tallies

Derby County20081112.95m

Sunderland2006154.025m

Watford200728-7.85m

Panel B: Three highest point tallies

Chelsea20059144.9m

Chelsea20069575.85m

Manchester Utd200897-14.8m

Source: O/S

Again, these outliers imply that success cannot be bought. Derby Countys respectable net spend of 12.95 million in the 2007-2008 season did not prevent them from achieving the Premier Leagues lowest ever number of points, while in the same season Manchester United won the Premier League with the most points in history, despite also recording a profit of 14.8 million in the transfer market.

Final proof of the effect of statistical outliers on the data can be produced by running the original regression without the observation for Chelsea in 2005. The regression, with 99 observations, becomes:Points = 50.27 + 0.188(total net spend)

Not only has the size of the marginal effect fallen quite dramatically, but, with a t-value of 1.36, the co-efficient is no longer statistically significant. The Chelsea 2005 observation, therefore, is playing a large role in producing a significant result for of 0.222.

The results for the second regression are of interest in that they show a clear and measured difference between the effectiveness of money spent in January and money spent in the summer. While summer spending brings about a noticeable, significant, and positive effect on performance, the marginal effect of January spending is barely positive and not significant. One interpretation of the large standard error for January spending is that it is a less accurate way of improving team performance, as players traded during the January transfer window cannot be relied upon to achieve their intended effect. Another, perhaps more plausible, interpretation of this results is that there is a selection bias in the teams that choose to engage actively in the January transfer market. Generally, a team which spends a lot in January is one which has performed in a manner below expectations in the first half of the season and is in need of investment to secure their objectives. This applies particularly to teams threatened with relegation. Tottenham, for example, spent 39 million in the January window of the 2008-2009 season in order to secure their Premier League status after a disastrous start to the season left the club in real danger of finishing in the bottom three.Conversely, there are several advantages to signing a player in the summer window. Firstly, for the majority of the summer window there are no league fixtures. This gives a new arrival the opportunity to settle down in the local area, integrate with team mates, and learn how his new team plays, without the pressure of competitive matches. Secondly, players signed during the summer window are more likely to have been thoroughly scouted and fit in with a managers long-term plans, rather than act as panic purchases to try to turn around a sinking ship. Thirdly, the new acquisition will participate in pre-season training with his new team, ensuring that he is fit and fresh for the start of the season.The differences in the characteristics of January and summer purchases are borne out in the regression. The effectiveness of summer spending is, on average, far greater than the effectiveness of January spending. There are two observations worth making at this point, however. The first is that the removal of Chelsea 2005 from the regression again renders spending insignificant. The second is that the upper bounds of the 95% confidence intervals for the two marginal effects are almost identical, at 0.52. This confirms that January spending is a lot more hit-and-miss in its effectiveness than summer spending.The high value on the co-efficient of total net spend from two seasons ago is among the most noteworthy of this paper. A marginal effect of .611 points for every 1 million of net spend two seasons ago implies a very clear and very large lag effect of spending. It signals that a certain bedding-in period exists between a player joining a club and that player affecting results. Moreover, the t-value on 3 is 3.58, which shows that it is highly significant. The co-efficients on 1 and 2 are both negative but have low t-values and are not considered significant.When the Chelsea data entry from the 2004/2005 season is removed from the regression, it might be expected that the co-efficient will become insignificant in the same manner that they did previously. In actual fact, the significance only drops slightly, to 3.54, demonstrating that, unlike the first two regressions, the Chelsea 2005 entry is not the major cause of the phenomenon. In fact, Chelseas two largest net spends over the period of study occur in 2004/2005 (75.85m) and 2005/2006 (44.9m) seasons, and are followed two seasons later by points tallies which are below Chelseas mean (2006/2007 83 points, 2007/2008 85 points). The fact that the results are robust in the absence of outliers affirms the effectiveness of lagged spending on performance.It is worth spending some time to think about the credible rationales behind this lagged marginal effect. It may take some time for a new player to acclimatise to his new surroundings, both sporting and non-sporting. It makes intuitive sense to believe that a player who is settled off the pitch is more likely play well on the pitch. This is particularly true of players arriving from foreign leagues, who may not be used to British life, the English language or the idiosyncrasies of the Premier League. Once these players are used to their new environment, it is easier for them to perform well for their new teams, and this acclimatization period may take a year or longer.The results of the regression may also be explained by money spent on young players, whose value at least as correlated with their potential talent as it is with their current talent. It may take time for expensive young players to break into a team or to match their potential with performances, and as a result it is possible that the marginal effect of expenditure on teenagers is lagged. Wayne Rooney, for example, joined Manchester United in the summer window of 2004 in a much-publicised 20 million transfer from Everton, but his effect on the team was a gradual one - the striker scored 17 goals in his first season, 21 in his second and 24 in his third. Evidently, as he developed as a player and settled in Manchester, his effectiveness for Manchester United increased. Theo Walcott, then aged 16, transferred from Southampton to Arsenal in the January 2006 window for a fee rising to 12 million, but over four years later has yet to establish himself as anything more than a bit-part player at the club. It can take time for young talented players to truly make their mark on the big stage. 5.3 Limitations of ResearchIn addition to transfers between clubs, there are other traditional mechanisms by which football clubs can acquire new players: free transfers and loans. The scope of this research did not extend to either, and this must be mentioned as a shortcoming of the results. A loan signing is a temporary arrangement between two clubs during which a player owned by one clubs joins, and plays for, the other club. Loan contracts tend to last for anywhere between a month and a year. At the end of the contract, the player returns to his parent club. Most loan deals take place for free, and in cases where a fee is paid, it is usually nominal. The 1995 Bosman ruling was a landmark case in sports law, and allowed players who had reached the end of their contract to move clubs, with no fee involved. This type of deal, known as a free transfer, is not overly common, especially at the top level of the game. However, several notable players have swapped clubs for no fee, including German captain Michael Ballack, who joined Chelsea from Bayern Munich in the summer transfer window of 2006.The omission of loan signings and free transfers from this research is significant in that, in terms this papers data, they represent an undocumented change in a clubs playing talent. Successful loans deals and free transfers will contribute to a teams points haul without showing up on their net spend. The most blatant example of this sort of effect during our period of study is the case of Argentinean Carlos Tevez. The highly rated striker arrived at West Ham United in August 2006 in unique circumstances, on loan not from his previous club (Brazilian side Corinthians) but a third party company which owned his sporting rights, Media Sports Investments (MSI). Tevez spent the 2006/2007 season at West Ham, and scored several important goals which were regarded as crucial in West Ham avoiding relegation that season. Tevezs arrangement with West Ham came to an end in the summer of 2007, and he then joined Manchester United on a similar loan deal from his third-party owners. He scored 19 and 15 goals respectively for Manchester United in their title-winning 2007/2008 and 2008/2009 seasons, but the club did not exercise their right to buy him, and Manchester City eventually purchased him from Media Sports Invetsments in summer 2009.While the Tevez case is an extreme one, it highlights how the exclusion of loans from the data diminishes its quality. Tevez had a relatively high profile effect on the outcome of the Premier League in first his two seasons in England, but is not included in any clubs net spend for those two seasons.

Outside of the transfer market, there are many other factors in the world of football which may contribute to an augmentation in a teams human capital. Two are noted here: youth systems and coaching. A clubs ability to develop and integrate graduates from their academy can act as a substitute to buying players in the transfer market, and may even lower net spend, through sales of home-grown talent. Similarly, good coaching and management can bring out the best from seemingly limited talent. In the framework of the regressions performed in this paper, both of these factors may have unobserved effects on points scored in the Premier League and consequently this must be considered a limitation of the research.The list of contributors to the success of a football team over the course of a season is vast, and could not possibly be covered in these pages. It would be simple-minded to believe that transfer spending is the only means towards achievement in football. A huge variety of variables play a role - the clubs manager, payroll, fixture schedule, injuries, internationals, fans, stadium, to name but a few. None of these variables are included in this papers regressions, despite their obvious impact on league performance.

The research also does not take into account the existing standards of the teams in 2004. The regression treats all teams as equal at the beginning of the period of study, a premise that clearly is not true. The existing state of each teams playing squad in 2004 is clearly a major constituent in determining the league winner in subsequent years, although naturally its influence diminishes with each passing season.Finally, the limitations of the data itself must be acknowledged. Although its source is relatively reputable, the size of transfer fees has not and cannot be verified by the parties involved. Instead, the data is treated as a best estimate of the true size of the values. The data set is also relatively small, limiting the significance of the data somewhat.5.4 Suggestions for further researchAn extension of the scale and scope of this research would provide further indication as to the salient inputs into a productive football team. A larger data set, including teams from leagues outside of England and studied over a period longer than five years, could produce more comprehensive and accurate data regarding the efficacy of transfer spending. A larger data set would also increase the accuracy of the co-efficient estimates and the confidence of the results. Given that the average net spend in the Premier League increased every season from 2004/2005 to 2008/2009, a consideration for football inflation could also be included in future research.A greater number of variables included in the regressions would present a clearer picture of the driving factors behind achievement in football. The low adjusted R-squared statistic for the first two regressions suggests that much of the variance in points attained in the Premier League is accounted for by factors other than net transfer spend. Some of the factors which might be included in a more complete regression are mentioned in Section 5.3, and in particular, a payroll variable and a variable which quantified the quality of a clubs manager might produce more significant results.

5.5 Conclusion

The finding that net transfer spend and league performance are positively correlated confirms not only the statistical relationship between money and success suggested by Hall, Szymanksi & Zimbalist (2002), but also the conventions of modern football management. A wider and more comprehensive set of measurements would reveal the direction and degree of causality in this relationship. The superior and more reliable effectiveness of summer spending over January spending reveals as much about the differing characteristics of the two as it does about the type of club that engages in each.

The primary conclusion to be drawn from this paper is that, although money is an important ingredient in Premier League, money alone cannot buy success. Chelseas Premier League title in the 2004/2005 season is an example of a large net spend achieving a large points tally, but even the spend alone does account entirely for their success. Instead, money acts as a complement to other pre-existing circumstances at a club, such as playing and managerial talent.References

1. Hall, Szymanksi & Zimbalist (2002) Testing Causality Between Team Performance and Payroll: The Cases of Major League Baseball and English Soccer Journal of Sports Economics 2002; 3; p. 149 2. Carmichael & Thomas (1995) Production and efficiency in team sports: an

investigation of rugby league football Applied Economics 1995; 27; p. 8593. Carmichael, Thomas & Ward (2000) Team Performance: The Case of English

Premiership Football Managerial and Economics Decisions 2000; 21; p. 31 4. Langelett (2003). The Relationship between Recruiting and Team Performance in Division Journal of Sports Economics 2003; 4; p. 2405. Leech & Barros 2006 (2006). Analyzing the Performance of the English FA Premier League with an Econometric Frontier Model Journal of Sports Economics 2006; 7; p. 3916. Soccer Base, http://www.soccerbase.com/(Chelsea 2005)

PAGE ii

_1332230400.unknown

_1332245226.unknown

_1332227741.unknown