What is the effect of UEFA Financial Fair Play on the competitive balance in European top football leagues? Name: Timo Leget Date: January 31, 2018 Abstract In this paper the effect of Financial Fair Play on the competitive balance in European top football leagues is analyzed, using two different measures of competitive balance. A time series analysis in order to forecast the competitive balance in the seasons after the start of Financial Fair Play (2011/12) is used to compare the forecasted values with the actual observed values. The results show that, for both measures, the C5 Index of Competitive Balance and Herfindahl Index of Competitive Balance, the competitive balance deteriorated after the start of Financial Fair Play. However, more research is needed to state whether these changes in competitive balance are solely explained by the FFP regulations. Other variables may also be responsible for the decrease in competitive balance. Name: Timo Leget Student number: 10770461 ECTS: 12 Supervisor: Vadim Nelidov
1 Statement of Originality This document is written by Student Timo Leget who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2 Introduction The past decades, European football clubs experienced booming revenues. The growing interest in the sport makes it interesting for investors looking for sponsorships and advertisement. According to the Benchmarking Report of the Union of European Football Associations (UEFA) (2011), the revenues of clubs participating in the first leagues in Europe has increased by 5.6 percent per year over the previous five years. For comparison, the revenues of the football clubs increased more than ten times faster than the overall economies of the European countries in this period. The total income (excluding transfer fees) of the clubs participating in the “big five” (England, France, Germany, Italy and Spain) increased from 4.2 billion to 8.4 billion euros. Despite the fast rising revenues, 63 percent of the European football clubs were operating at a loss in 2011. On top of that, the aggregated losses of European football clubs have tripled since 2007 to 1.7 billion euros (Serby, 2014; Storm & Nielsen, 2012). This trend of rising revenues and increasing losses forced the UEFA to impose regulations since 2011 in order to keep clubs financially healthy and protect from bankruptcy. This regulation is called UEFA Financial Fair Play (FFP), which will be more extensively explained later on in this thesis. In short, FFP makes sure that European clubs are no longer able to overspend on their budgets (Sass, 2012). Other than in ‘normal’ sectors, sport teams jointly benefit from a strong competitive balance in their league. An important property to any sport is the idea that “competitors must be of approximate equal ‘size’ if any are to be successful” (Rottenberg, 1956). In this light, it seems reasonable for the UEFA to impose financial regulations on European clubs, in order to prevent these clubs from financial difficulties and to ensure ‘strong’ competition. However, this regulation gained a lot of criticism among economists. They argue that imposing financial regulations on European clubs could have negative consequences for the competitive balance in the leagues (Lindholm, 2010; Sass, 2012; Sass 2014; Szymanski 2014; Vöpel, 2013). On the other hand, some economists argue that FFP leads to more efficient management and is an improvement for the competitive balance in the leagues (Franck, 2014; Muller et al., 2012; Kesenne, 2006). Clearly, there is no consensus in the literature about the effects of FFP and it still leads to a lot of debate. This thesis will attempt to contribute to this
3 debate, resulting in the following question: What is the effect of UEFA Financial Fair Play on the competitive balance in European top football leagues? The remainder of this thesis is structured as follows. In the following section, the literature review will be presented. First, I will explain what FFP exactly is. Consequently, the different opinions of economists on FFP regarding the concept of competitive balance will be presented. The next section contains the methodology of this thesis, in this section will be explained which measures will be used and what analysis will be done. In the fourth section, the results of the analysis are presented and in the last section I will show my conclusion and discussion.
Literature Review In this section will be discussed what FFP exactly is and the concept of competitive balance will be more extensively explained. Consequently, the different opinions of economists on FFP with regard to the idea of competitive balance will be analyzed and finally the hypothesis of this thesis will be stated. According to the official website of the UEFA, FFP is about “improving the overall financial health of European club football” (UEFA, 2015a). The objectives of the UEFA by implementing FFP are to protect long-term financial stability of European Club Football and restoring competitive balance between clubs and leagues (Vöpel, 2011). The UEFA introduced FFP in 2011 as an extra requirement in order to get a license as professional football club. The exact rules of the regulation are stated in the UEFA Club Licensing and Financial Fair Play Regulations (2015b). The regulation is based on two major financial requirements. The first requirement is that clubs have to prove that they: “do not have overdue payables towards other clubs, their players and social/tax authorities throughout the season” (UEFA Club Licensing and Financial Fair Play Regulations,2015). In other words, clubs have to prove they have paid their bills. The second requirement is called the break-even requirement, which is an obligation for the clubs to balance their revenues and expenses over three consecutive years on average. Furthermore, the UEFA makes a distinction between cash flows from football and non-football related
4 activities. In order to pass the break-even requirement, clubs have to balance their football related cash flows, which are called the ‘relevant’ revenues and expenses (Serby, 2014). Art. 58 of the FFP Regulations states: “Relevant income is defined as revenue from gate receipts, broadcasting rights, sponsorship and advertising, commercial activities and other operating income, plus either profit on disposal of player registrations or income from disposal of player registrations, excess proceeds on disposal of tangible fixed assets and finance income. It does not include any non-monetary items or certain income from non-football operations.” In short, it is no longer allowed to use income from private ownership investments to match the football related expenses (e.g. buying players). However, as stated in Art. 61, an “acceptable deviation” of 5 million euros or up to 30 million euros (in case of full coverage by private investor) over the previous three seasons is allowed. The regulation is backed by eight sanctions, in case of breaking one of the requirements. These sanctions, in order of severity, are: reprimand, warning, fines, points deduction, withholding of revenue from a UEFA competition, prohibition to register new players for UEFA competitions, restrictions on how many players a club can register for UEFA competitions, disqualification from a competition in progress and exclusion from future competitions (UEFA.com, June 2015). Imposing budget restrictions on professional sport teams is not unusual. These restrictions have been regularly imposed in the US, always with the intention to restore competitive balance. European courts have decided that restoring competitive balance is a legitimate reason to impose budget restrictions in sports (Peeters & Szymanski, 2014). For this reason, it is useful to have a clear idea of the concept of competitive balance. Many sport economists referred to this concept. The main idea is that, contrary to ‘usual’ economic theory, the product of sports is a joint-production of the teams within a league, dependent on the effort and cooperation of every team (Morrow, 2003; Goossens 2006; Lee and Fort, 2012; Freestone & Manoli, 2017). Neale (1964) already mentioned that in sport leagues “a pure monopoly is a disaster” in order to develop an attracting league. Moreover, according to Michie and Oughton (2004) a lack of competitive balance in a league also results in other risks than declining interest. For instance, the possibility of
5 bankruptcy of lagging clubs increases. The second risk that arises is the search for competitive balance for top teams in a new league, which will ultimately weaken the national leagues of the remaining clubs (Michie & Oughton 2004). The competitive balance in European football has been especially decreasing since the turnover of the century, partly caused by the enormous increases in prize money for participating in the UEFA Champions League (Curran et al, 2009; Lee & Fort, 2012). Over the last decade, 50 percent of the Champions League prize money has been earned by just ten different clubs. This unequal distribution of prize money causes a decline in competitive balance, since the clubs participating in the Champions League are able to use this money in order to gain a better competitive position in their national league (Szymanski, 2014). Goossens (2006) observed that a lot of discussion is going on how to measure competitive balance in sport leagues. Which is supported by the statement of Zimbalist (2002): “there are always as many ways to measure competitive balance as there are to quantify money supply”. In the methodology will be discussed what measures will be used in this thesis. Despite the unanimous approval of implementing FFP, a lot of criticism raised with regard to the regulations. Critiques had many different perspectives: the perception that FFP will lead to a decrease in quality of the teams (Madden, 2012); the concern that de regulations will have unintended downward pressures on the salaries of players (Preuss et al., 2014); the legality of the regulations (Peeters & Szymanski, 2014; Lindholm, 2010). However, the majority of the criticism centered around the belief that FFP regulations will have negative consequences for the competitive balance. One of the major concerns with regard to the FFP regulations is explained by Dietl et al. (2009). In this paper is proved that initial endowments affect initial investments. The clubs with lower initial endowments invest more in the first period than their competitors with higher endowments in order to fix the competitive gap (Dietl et al., 2009). In this sense, it would be unreasonable to limit clubs in their investment opportunities, since it will reduce the possibility for smaller teams to catch up with the bigger teams, which is positive for competitive balance. Markus Sass (2012) analyzed the long-term development of competitive balance in a league facing spending limits, as is the case with FFP. His model shows only one equilibrium, in which
6 the smaller teams are totally dominated by the bigger teams of a league and competitive balance is maximally uneven (Sass, 2012). The concept behind this ‘status quo’ is that when a club becomes more successful, it is able to attract more money via spectators, advertising, sponsorships and so on. This will increase the relevant income, which will be used to buy better players and will probably lead to new successes. In short, a vicious circle of ‘money generates money’ is originated (Sass, 2012). According to Lindholm (2010), the new regulations are likely to decrease competitive balance, since FFP regulations are an indirect form of a salary cap. This means that bigger clubs will have big competitive advantages over the smaller teams, since they have much higher salary budgets. On top of that, the deflating effect on salaries may lead to a reduction of the quality of the players in a league. Top European players will look for better options beyond Europe, and new talent will be reluctant to enter the European football market. A downward pressure on salaries thus might lead to deflationary effects on the quality of the players. Since the bigger teams are more likely to keep their salary budgets high, this decrease in quality of the players will affect smaller teams the most, which will decrease competitive balance (Budzinski & Feddersen, 2014). Preuss et al. (2014) approached the possible decline in competitive balance building a simple game theory model. Their analysis proves that, because of the presence of a Prisoner’s Dilemma, clubs have the incentive to bypass the FFP regulations. This causes costs for both the clubs and the UEFA. In addition, smaller clubs may not have the resources to participate in this ‘bypass-game’, since they have not the budget for hiring legal experts. Eventually, the wealthier teams will end up with a competitive advantage and thus the league will be more unbalanced. Peeters and Szymanski (2014) proved, using a data analysis, that the break-even requirement does not improve competitive balance and protects the market share of traditional top teams, since it reduces the opportunities of smaller teams. On top of that Peeters and Szymanski argue that the UEFA is (partly) responsible for the increasing competitive unbalance, since it decides how the prize money of the European tournaments is distributed among the participants.
7 Nevertheless, Franck (2014) argues that FFP increases the incentives for good management and innovation, which could help to dismantle the established competitive unbalance. Moreover, limiting clubs to their own spending power will help to rebalance the competition, since bigger teams are no longer able to gain advantages with help from private investments (Franck, 2014). Sport leagues have in common that the participants are seeking for win maximization, according to Kessene (2006). This results in a ‘rat race’ for the best players and forces managements of clubs to make irrational decisions. In this point of view, the FFP regulations could bring rationality in the managements back in European football, which will ultimately result in a better balanced league (Kesenne, 2006). Furthermore, Muller, Lammert and Hovemann (2012) showed that FFP regulations have impact on the use of financial doping. Financial doping is explained as “performance-oriented financial means not earned by a club directly or indirectly through its sporting operations or drawing potential, but rather provided by an external investor, benefactor, or creditor detached from sporting merit and drawing potential as well as from sustainable investment motivations” (Muller et al., 2012). Financial doping shifts competition from the pitch to the quest for the wealthiest investor. This shift of competition lowers the attractiveness of the sport, which makes it justified to regulate financial doping in FFP regulations. Since, excessive funding is no longer allowed to use for football related activities, Muller et al. argue that FFP limits the possibilities of financial doping. Hence, the competitive balance will be less affected by financial doping (Muller et al., 2012; Ernst & Young, 2010). Since the FFP regulations are recently implemented, finding literature which actually tests the impact of FFP on competitive balance in European football is difficult. However, one paper did research on this topic. Freestone and Manoli (2017) conducted a research using two three different measures of competitive balance. All of these measures are based on the points total of each team at the end of the season (Freestone & Manoli, 2017). Their research solely analyzed the English Premier League in the period from 1992/93 up to 2015/16. All three measures indicate a general decline in competitive balance in the EPL since 1995/96, which is
8 exactly the period the UEFA started to increase the prize money of the Champions League (Freestone & Manoli, 2017). Whilst the measures used by Freestone and Manoli for competitive balance in the EPL showed increasing competitive unbalance the first three years after the start of FFP, the two subsequent seasons indicated improvements for competitive balance (Freestone & Manoli, 2017). Finally, Freestone and Manoli conclude that the data suggests that there is little evidence to support the criticisms against FFP regarding competitive balance. However, the last two seasons analyzed, showing positive trends for competitive balance, suggest the opposite and FFP is having positive impact on competitive balance. Nevertheless, Freestone and Manoli mention that there is currently not enough data to fully support one of the claims (Freestone & Manoli, 2017). This literature review shows that there is no consensus among economists whether FFP has positive or negative effects on the competitive balance, if it has any effect at all. This debate is exactly what this thesis makes worth writing. The research that has been conducted by Freestone and Manoli is an interesting starting point. This is why I follow their, preliminary, conclusion for my hypothesis: UEFA Financial Fair Play did not have impact on competitive balance in European top football. Methodology In this section, I will first explain the theory behind measuring competitive balance, which results in the two measures used for this thesis. After explaining how the two different measures work for this thesis, the way of collecting the data will be explained. This research will use a time series analysis, where the predicted competitive balance after the start of FFP will be compared with the actual competitive balance. In the last part of this section will be explained what time series is and how this will be used regarding the competitive balance and FFP. Competitive balance is a measure of the relative performance of a team within a particular sports league, used to give an indication of how evenly matched the teams in the league are (Goossens, 2006; Freestone & Manoli, 2017). Competitive balance is important for all professional sport leagues, because of its impact on demand (Neale, 1964). In the past years, this impact has been even more important. Nowadays, it is not just ticket sales at league games,
9 but also broadcast rights, which can be sold globally and represents the demand of the fans for the sport (Manasis et al., 2015). Because of the importance, competitive balance is a prominent subject in sports economics. However, as with a lot of economic concepts, competitive balance is a latent variable which cannot be measured directly (Manasis et al., 2015; Freestone & Manoli, 2017; Goossens, 2006; Michie & Oughton, 2004). The intention of this thesis is to measure the competitive balance in European top football, which, according to Gerrard (2004) is the “heartland of football”. However, European football leagues are complex in structure. European national leagues usually offer multiple prizes (championships and qualifications for Champions League or Europa League), this set up of leagues is a big difference compared to the traditional single prize league system used in the United States (Kringstad & Gerrard, 2007). According to Freestone and Manoli (2017) the appropriate measures of competitive balance depend on the rules, structure and the format of the league. This thesis will use measures based on the points totals of the teams at the end of every season, since this metric provides the most accurate signal of overall performance in the league (Evans, 2014). The leagues considered in this thesis make use of the same points-system, teams are awarded with zero points for a loss, one point for a draw and three points for a victory. Nevertheless, points are not the only way to measure competitive balance. Win percentages, goals scored, goal differences, goals conceded and league position can all be used as metric to measure competitive balance. However, the win percentage, often used in American sports leagues, does not make a distinction between a draw and a loss, which makes it not fully compatible for measuring competitive balance in European football (Quirk & Fort, 1992). The metrics suggested using goals, produces useful information regarding the relative performance of a team, but using it as single indicator for competitive balance may lead to wrong conclusions. Points, and therefore prizes, are not won by the the total amount of goals scored and conceded in a season, but instead are determined by goals every single match (Evans, 2014; Freestone & Manoli, 2017). On top of that, some teams may use different tactics which affects the amount of goals scored and conceded, but may result in higher points won in
10 a season. Finally, the suggested metric using league position as a measure of competitive balance may also lead to wrong conclusions, since it provides no useful information about the relative differences between the participating teams (Freestone & Manoli, 2017). Considering that the league position at the end of the season is determined by the total amount of points a team has won, the team with the highest amount of points wins the league. In this light, it makes sense to use measures based on the same metric (Freestone & Manoli, 2017). According to Evans (2014), it is the points-based system that best captures the overall and relative team within a league of this nature. In similar papers, the authors use two or more measures to make sure that the results are robust and not due to the choice of a specific measure (Freestone & Manoli, 2017; Goossens, 2006; Brandes & Franck, 2007; Michie & Oughton, 2004). This thesis will adopt this approach of different measures and will use two different measures. The first method, the five club concentration ratio and index of competitive balance, measures the extent to which a league is dominated by the five highest ranked clubs. In other words, it measures the inequality of the top five and the rest of the league. The five club concentration ratio (C5) can be calculated according to the following formula (Michie & Oughton, 2004; Koning, 2000):
C5 Ratio =
𝒕𝒐𝒕𝒂𝒍 𝒑𝒐𝒊𝒏𝒕𝒔 𝒘𝒐𝒏 𝒃𝒚 𝒕𝒉𝒆 𝒕𝒐𝒑 𝒇𝒊𝒗𝒆 𝒄𝒍𝒖𝒃𝒔𝒕𝒐𝒕𝒂𝒍 𝒑𝒐𝒊𝒏𝒕𝒔 𝒘𝒐𝒏 𝒃𝒚 𝒂𝒍𝒍 𝒄𝒍𝒖𝒃𝒔=
𝟓𝒔
𝒊 𝒊4𝟏where 𝒔𝒊 is the share of points of the club i. The C5 ratio depends on the number of clubs in the league and the degree of inequality between the top 5 clubs and the rest of the league. An increase in the C5 ratio reflects a decrease in the competitive balance and more domination by the top 5 clubs. Since it is impossible for the top 5 clubs of the league to win all the points, the boundaries of the index lie between 5/N, where N is the number of participating clubs, and M/(M+T), where M is the maximum number of points attainable by the top 5 clubs and T is the minimum number of points the rest of the participating teams could end up with (Michie &
11 Oughton, 2004). According to Michie and Oughton (2004), the biggest advantage of using this formula is that it is easy to understand. On top of that, Michie and Oughton (2004) observe that the C5 ratio reflects neatly one of the factors responsible for declining competitive balance. Since the top 5 clubs in a league regularly play in European club tournaments, leading to more money for these clubs, the national competitive balance is disturbed. However, changes in league size affect the competitive balance of the league significantly. In order to correct for these changes in league size the following index is composed by Michie and Oughton (2004):
C5 Index of Competitive Balance= (
𝟓/𝑵𝑪𝟓)*100
where N is the number of clubs in the league. For a perfectly balanced league the index has a value of 100. Decreases in competitive balance will lead to an increase in the C5 index of competitive balance (C5I). This thesis will try to combine the leagues from different countries (with different league size), which makes it justified to use the C5I instead of the C5 ratio without adjustment for league size. A disadvantage of the method described above is that it does not capture changes in imbalances within the top five clubs or within the bottom (rest) of the league. The second measure used in this thesis will be the Herfindahl Index of Competitive Balance (HICB). The Herfindahl Index, developed by Herfindahl, looks at inequalities between all the firms in an industry. Applying to football, this means that the Herfindahl Index captures inequalities between all the clubs participating in a league (Freestone & Manoli, 2017; Brandes & Franck, 2007; Goossens, 2006; Michie & Oughton, 2004). The Herfindahl Index puts focus on each club’s share of points and aggregates these into an index using each club’s share of points as weights (Brandes & Franck, 2007):𝐇 =
𝑵𝒔
𝒊𝟐 𝒊4𝟏12 where s is club i’s share of points in a season, and i = 1, 2, ...N, where N is the number of clubs in the league. An increase in the index indicates an increase in inequality, thus a decrease in competitive balance. Like the before discussed C5 ratio, changes in the league size affect the Herfindahl index. This can be corrected by dividing H by the value of H that would be attained in a perfectly balanced league. This results in the Herfindahl Index of Competitive Balance (HICB) (Michie and Oughton, 2004):
HICB = (
𝟏/𝑵𝑯)*100
where H is the former discussed Herfindahl index and 1/N is ∑𝑝?@ where pi equals the share of points won by club i in a perfectly balanced competition of any size. A perfectly balanced league would be indicated by a HICB-value of 100 and a decrease in competitive balance is expressed by an increase in the HICB (Freestone & Manoli, 2017; Brandes & Franck, 2007; Goossens, 2006; Michie and Oughton, 2004). The data needed for this thesis are the end of season league tables of the leagues included in this thesis. These league tables are openly available on the websites of the national football associations. This thesis will analyze the five biggest competitions of Europe (Premier League, England; La Liga, Spain; Bundesliga, Germany; Ligue 1, France; Serie A, Italy) from the season 1992/93 up to 2016/17. Beginning in the season 1992/93 makes sense, since in this year the Premier League in England has been reformed and this is seen as the start of the ‘big money coming into European football’ (Freestone & Manoli, 2017). After applying the two described measures for the five leagues separately, this thesis will combine the data of the five leagues into one data set based on the average of the five leagues. Using this data set makes it possible to draw a more general conclusion for European football. Once the data is collected, a time series analysis will be conducted. Time series can forecast the development of the value of a variable in the future (Stock & Watson, 2012). This thesis will forecast, using time series analysis, the values of competitive balance (C5I and HICB) for the seasons after the start of FFP in European football. Since the actual data on competitive balance in these seasons (until 2016/17) are known, time series analysis and forecasting makes it possible to compare the forecasted and actual values of competitive balance (Stock &13 Watson, 2012). If the forecasted and actual values of competitive balance in the seasons 2011/2012 up to 2016/2017 show much difference, this could be evidence of the impact of FFP on competitive balance in European football. The time series model that will be used in this thesis is called the 𝑝AB-order autoregressive model, or AR(p) model. Forecasts using an autoregression relate a time series variable (C5I and HICB) to its past values. The AR(p) model represents 𝑌A as a linear function of p of its lagged values, that is in this model 𝑌ADE ,𝑌AD@, . . , 𝑌ADH, plus an intercept. The number of lags, p, included in the AR(p) model is called the lag length or the order of the autoregressive. This leads to the following formula (Stock & Watson, 2012):
𝑌
A= 𝛽
J+ 𝛽
E𝑌
ADE+ 𝛽
@𝑌
AD@+ ⋯ + 𝛽
H𝑌
ADH+ 𝜇
Awhere 𝜇A is an error term.
In practice, 𝛽 is unknown, so forecasts must be based on estimates of 𝛽. This thesis will use the OLS estimators 𝛽, which are constructed using historical data. In general,𝑌NOE|N will denote the forecast of 𝑌NOE, based on information through period T using a model estimated with data through period T. Accordingly, the forecast based on the AR(p) model is:
𝑌
NOE|N= 𝛽
J+ 𝛽
E𝑌
N+ 𝛽
@𝑌
NDE+ ⋯ + 𝛽
H𝑌
NDHOEThe forecast will be presented with a 95% confidence interval, which will be used to judge whether FFP affects the competitive balance. In the case that the actual observed values exceed the 95% confidence interval, this might be evidence for the impact of FFP on the competitive balance in the top football leagues in Europe.
14 Data This section will discuss the obtained data, using the measures for competitive balance. In the appendix, two tables are presented with a complete overview of the data. The first table presents the data regarding the C5 Index of Competitive Balance and the second table shows the data corresponding with the Herfindahl Index of Competitive Balance. All leagues are observed for 25 seasons in a row and the averages of the five leagues per season are presented on the right side of the tables. According to the C5I, the Premier League is the league with the most competitive unbalance on average, with a mean of 142.0247. The other four leagues are slightly more balanced with averages between 140.3573 (Serie A) and 133.7509 (Ligue 1). The highest competitive unbalance average in the Premier League was expected. Freestone & Manoli (2017) already called the Premier League the most unbalanced league in European football. The highest standard deviation of the C5I is found in the Spanish La Liga (8.731857) and the lowest standard deviation is observed in the Serie A (5.778707). The most unbalanced season, measured with C5I, in these 25 seasons, is observed in the Spanish La Liga with 159.0086 (2014/15) and the most balanced season with 121.1456 is observed in the French Ligue 1 (1999/00). Considering the HICB, the most unbalanced league on average is the Italian Serie A, with a mean of 109.458. The most balanced league is the French Ligue 1, with an average of 106.5152. The league with the highest standard deviation, considering HICB, is again the Spanish La Liga (3.109332) and the lowest standard deviation is found in France (1.811999). The highest and lowest observed competitive unbalance were again the 2014/15 and 1999/00 in Spain (114.9599) and France (102.8566). The remainder of this thesis will be done using the average of the measures over the five leagues.
15 Analysis In order to be able to conduct time series analysis, the data has to be tested for some properties. First of all, the data has to be tested for stationarity, which is the case if a time series’ probability distribution does not change over time. Stationarity requires the future to be like the past, at least in a probabilistic sense (Stock & Watson, 2012). Consequently, it has to be decided how many lags will be included in the model, using information criteria. Lastly, the results from the AR(p) model will be presented and a forecast will be conducted. Brandes and Franck (2007) tested for stationarity in a comparable research, using the Augmented Dicky-Fuller Unit Root Test IADF). Under the null hypothesis for the ADF, 𝑌A is nonstationary and under the alternative hypothesis 𝑌A is stationary. Before being able to carry out the ADF, it should be considered whether 𝑌A is subject to a linear time trend. If so, this trend must be added to the ADF (Stock & Watson, 2012). Figure 1: Average C5I Before FFP Since this thesis will compare the forecasted values for the years after the start of FFP with the actual observed values, at this point the analysis only needs the observations up to the start of FFP. Figure 1 presents the development of the average C5I before FFP, where the 128 130 132 134 136 138 140
Average C5I Before FFP
Average C5I Before FFP Lineair (Average C5I Before FFP)16 dashed line represents the trend over the years. Clearly, there is a positive trend observable, which means that this should be taken into account when carrying out the ADF. A positive linear trend is also observed in the case of the HICB, as presented in figure 2: Figure 2: Average HICB Before FFP I continue with the ADF itself, in order to check the data for stationarity. As shown in table 1, the test statistic for both the C5I and the HICB exceed the one percent critical value (-4.380). The p-value for the C5I is 0.0001 and for the HICB is 0.0005, which means that the null hypothesis can be rejected and we may assume the data to be stationary. Table 1: ADF Test
Augmented Dicky-Fuller Test Measure Test Statistic 1% Critical Value 5% Critical Value
C5I (with trend) -5.308 (0.0001) -4.380 -3.600 HICB (with trend) -4.791 (0.0005) -4.380 -3.600 105 106 107 108 109 110
Average HICB before FFP
Average HICB Before FFP Lineair (Average HICB Before FFP)17 After checking for stationarity, the next step is to decide how many lags should be included in the AR(p) model, as described in the methodology. This decision can be based on various so-called information criteria, which is done by Brandes & Franck (2007) in their paper. Simply stated, information criteria are based on the error terms’ variance, i.e. the unexplained part of the model. Thus, the lag structure yielding the lowest information criterion is the best (Lee & Fort, 2012; Brandes & Frank, 2007). The best known information criteria are the Akaike information criterion (AIC), Hannan-Quinn information criterion (HQIC) and the Schwarz criterion (SIC). In this thesis, the SIC, also known as the Schwarz Bayesian Information Criterion (SBIC), will be leading to decide on the amount of lags included in the autoregressive model. The reason for this choice is motivated by the fact that SC generally performs better in an analysis with a small sample size, which is the case in this thesis (Lee & Fort, 2012; Brandes & Franck, 2007; Lütkepohl, 2006). However, for reasons of completeness the other two mentioned information criteria will also be presented. Table 2: Lag decision C5I Table 3: Lag decision HICB As shown in the tables 2 and 3, all three information criteria have the least error term variance if two lags are included in the autoregressive model. This thesis will follow these information criteria and will include in both the C5I- and HICB-analysis two lags.
Lag decision C5I SBIC AIC HQIC
L1+L2 4.601 4.454 4.468
L1+L2+L3 4.810 4.617 4.627
L1+L2+L3+L4 5.074 4.838 4.835
Lag decision HICB SBIC AIC HQIC
L1+L2 3.159 3.012 3.026
L1+L2+L3 3.284 3.091 3.101
18 As presented in table 4, which shows the results of the AR(2) model regarding C5I, only the second lag is significant in this model (p-value: 0.001). However, excluding the insignificant first lag (p-value: 0.528) resulted in a slight decrease in the information criteria from 4.601 to 4.457 (SBIC) and a slight decline in the R-square from 0.424 to 0.411, Considering these results, I decided to keep the first lag in the model in order to maintain the (slightly) higher R-square. Table 4: Output AR(2) model for C5I C5I In table 5, presented on the following page, the results for the HICB are shown. This model shows again significance for only the second lag (p-value: 0.005). However, excluding the insignificant first lag (p-value: 0.959) results in a slightly lower error variance (change of 0.167) and the R-square practically does not change (change of 0.0001). In that sense, it does not make much difference to exclude the first lag from the model. Since I prefer to keep both models as equal as possible, I decide to keep the model with two lags and thus choose for a slightly higher error variance. In short, for both models is decided to include the first two lags.
C5I Coef. Std. Err. t-value P>t 90% LB 90% UB
L1. 0,117192 0,2045518 0,57 0,576 -0,2430872 0,4774713 L2. 0,6932336 0,2305942 3,01 0,009 0,2870857 1,099381 const. 25,69195 37,299130 0,69 0,502 -40,00339 91,38728 Number of obs 17 AIC 4,453847 HQIC 4,468462 SBIC 4,600884 R-squared 0,4240 Prob>F 0,0210 RMSE 2,07216
19 Table 5: Output AR(2) model for HICB
HICB Coef. Std. Err. t-value P>t 90% LB 90% UB
L1. 0,0106542 0,2278496 0,05 0,963 -0,3906597 0,411968 L2. 0,6265469 0,2447437 2,56 0,023 0,1954773 1,057616 const. 39,1723 29,74425 1,32 0,209 -13,21655 91,56115 Number of obs 17 AIC 3,011779 HQIC 3,026395 SBIC 3,158816 R-squared 0,3449 Prob>F 0,0518 RMSE 1,00759 After testing for stationarity and deciding on the lags including in the model, the next step is to actually forecast the competitive balance (C5I and HICB) for the seasons after the implementation of FFP. These forecasted seasons are from 2011/12 up to 2016/17 and will be compared with the actual observed values for these seasons. The forecasted values will be accompanied by a 90% confidence interval. If the actual observed values of competitive balance exceed the 90% confidence interval, one could assume that some factor has impacted the competitive balance in these seasons. Table 6: Forecast, Confidence Interval and Actual values C5I
Season Forecast C5I Lowerbound Forecast C5I Upperbound Forecast C5I Actual value C5I
2011/12 137,246 133,559 140,953 137,479 2012/13 135,396 131.680 139,112 139,202 2013/14 136,710 131,945 141,474 148,058 2014/15 135,574 130,326 140,823 144,146 2015/16 136,352 130,904 141,800 142,326 2016/17 135,656 130,163 141,149 149,389
20 Figure 3: Forecast C5I vs Actual C5I As presented in table 6 and figure 3, in all seasons after the start of FFP the actual observed value of C5I was higher than the forecasted value in the corresponding season. On top of that, in five out six seasons the actual observed value exceeded the upperbound of the 90% confidence interval. This means that the competitive balance, measured with C5I, has decreased more than expected in the years after the start of FFP, which is consistent with the literature of the economists arguing that FFP would have negative effects on the competitive balance. However, concluding that solely FFP affected the competitive balance negatively is one step too far right now. There may be some other variables that have been changed over the years that impacted competitive balance as well. Table 7: Forecast, Confidence Interval and Actual values C5I 125 130 135 140 145 150
Forecast C5I vs Actual C5I
Actual C5I Forecast C5I Lowerbound Forecast C5I Upperbound Forecast C5I
Season Forecast HICB Lowerbound Forecast HICB Upperbound Forecast HICB Actual value HICB
2011/12 109,006 107,208 110,804 109,185 2012/13 107,371 105,592 109,150 110,019 2013/14 108,614 106,429 110,800 111,784 2014/15 107,603 105,282 109,923 109,897 2015/16 108,37 106,034 110,710 109,758 2016/17 107,745 105,415 110,075 112,277
21 Figure 4: Forecast HICB vs. Actual HICB Table 7 and figure 4 regarding the HICB, show the same trend as the results for the C5I. In all seasons after the start of FFP, the actual observed values of HICB all exceed the forecasted values for the same seasons. In contrary with the case of the C5I, the actual observed values of the HICB exceed in half of the seasons the 90% confidence interval. This means that the competitive balance, measured with HICB, has decreased in some cases after the start of FFP. Which is an indication that FFP affects the competitive balance of European football, according to the HICB. However, since the competitive balance has not exceeded the confidence interval in the other three seasons, it is premature to conclude that the competitive balance definitely decreased after the start of FFP. Nevertheless, this analysis for the HICB suggests a decreasing trend of competitive balance after the start of FFP. In five out of the six seasons after starting FFP, the actual observed value of competitive balance was higher than the highest observed value of competitive balance in the seasons before FFP. Again, we should be careful with concluding that this suggested trend of deteriorating competitive balance is due to FFP, there may be other factors affecting the competitive balance. Examples of factors affecting the competitive balance, for both HICB and C5I, are increasing 105 106 107 108 109 110 111 112 113
Forecast HICB vs. Actual HICB
22 transfer expenditures, unequal distribution of broadcasting rights and increasing prize money for European club competitions. These factors might also be responsible for the differences observed between the C5I and HICB. The variables mentioned are all dealing with financial developments. It is known that the top clubs in a league are more financially rewarded than the teams ending at the bottom of the league. Especially, European prize money (Champions League and Europa League) is increasing the past years, which has consequences for the equality in a national league. Only four or five teams from a league qualify for a European competition every year, which means more income for these clubs comparing to the rest of the league. It is possible that the teams generating more income from European competitions are able to keep and strengthen their top five position in a league more easily, which leads to less competitive balance. On top of that, clubs participating in the Champions League are more interesting for advertisers and sponsors, which leads to more income for these clubs. These developments lead to more financial inequality and less competitive balance within a league. Since the C5I measures the inequality between the top five clubs and the rest of a league, it makes sense that this measure is more affected the past years than the HICB measure. The inequality between the top five teams and the rest of the league probably has grown faster than the inequality between all teams, which is measured by the HICB. Conclusion In this thesis I analyzed the impact of the FFP regulations implemented by the UEFA on the competitive balance in European top football. The reason for this research is the strong debate among economists. One side argues that the FFP regulations negatively affect the competitive balance of European football, where the other side’s view is that FFP increases the competitive balance. For this thesis, I used two different measures. The C5I measures the inequality between the top 5 teams of the league and the rest of the league. And the HICB measures the inequality between all clubs in a league. After collecting the data, I conducted a time series analysis using the AR(p) model.
23 Checking for stationarity and deciding on the including lags made it possible to forecast the competitive balance for the seasons after the start of FFP. Consequently, this thesis compared the actual observed values of competitive balance with the forecasted values, which showed a positive trend; in all seasons exceeded the actual value the forecasted values, which is a sign of decreasing competitive balance. For the C5I analysis, in five out of the six seasons after the start of FFP the actual value exceeded the upperbound of the confidence interval, which may be an indication that FFP affected the competitive balance negatively. In the case of the HICB analysis, in half of the seasons after the start of FFP the actual value exceeded the upperbound of the confidence interval. In five out of the six seasons after starting FFP, the actual observed value of competitive balance was higher than the highest observed value of competitive balance in the seasons before FFP which also suggests a negative impact on the competitive balance due to FFP. However, I am not able to fully conclude that FFP negatively impacted the competitive balance in European top football. Other variables may also be responsible for the deterioration of competitive balance, which are not included in the model. This thesis contributes to the debate on the effect of FFP on the competitive balance in European football. However, more research has to be done in order to able to draw more firm conclusions. For future research, there are some improvements on this thesis. First of all, the sample size in future research should be enlarged. The more seasons can be analyzed, the stronger the conclusion would be. However, since the relatively short period after the start of FFP, only patience will increase the sample size. On top of that, including some extra explanatory variables would be helpful in order to improve the predictable power of the models. For example, data on Champions League prize money and financial data may be interesting to include in the models in the future. However, including these variables is beyond the scope of this thesis.
24 References Brandes, L. and Franck, E. (2007). Who made who? An Empirical Analysis of Competitive Balance in European Soccer Leagues. Eastern Economic Journal, 33(3), 379-403. Budzinski, O. and Feddersen, A. (2014). The Competition of Economcs of Financial Fair Play. Contemporary Research in Sports Economics, 77-98. Bundesliga. (2017). “Tables”, available at: https://www.bundesliga.com/en/stats/table/, (accessed November, 2017). Curran, J., Jennings, I. and Sedgwick, J. (2009). Competitive Balance in the Top Level of English Football, 1948-2008, an Absent Principle and a Forgotten Ideal. The International Journal of the History of Sport, 26(11), 1735-1747. Dietl, H., Franck, E., and Lang, M. (2009). Overinvestment in Team Sports Leagues: a Contest Theory Model. Scottish Journal of Political Economy, 55(3), 353-368. Ernst & Young. (2010). Football Meets Finance VII. London, UK. Evans, R. (2014). A Review of Measures of Competitive Balance in the ‘Analysis of Competitive Balance’ literature. Birkbeck Sport Business Centre Research Paper Series, 7(2), 1-59. Franck, E. (2014). Financial Fair Play in European Club Football – What is it all about? International Journal of Sport Finance, 9(3), 193-217. Freestone, C. J. and Manoli, A. E. (2017). Financial Fair Play and Competitive Balance in The Premier League. Sport, Business and Management: An International Journal, 7(2), 175-196. Gerrard, B. (2004). Still Up for Grabs? Maintaing the Sporting and Financial Viability of European Club Soccer. International Sports Economics Comparisons, Praeger, Connecticut, London. Goossens, K. (2006). Competitive Balance in European Football: Comparison by Adapting Measures: National Measure of Seasonal Imbalance and Top 3. Rivista di Diritto ed Economia dello Sport, 2(1), 77-122.
25 Kesenne, S. (2006). The Win Maximization Model Reconsidered: Flexible Talent Supply and Efficiency Wages. Journal of Sports Economics, 7(4), 416-427. Koning, R.H. (2000). Balance in Competition Dutch Soccer. Journal of Royal Statistical Society: Series D (The Statistician), 49(3), 419-431. Kringstad, M. and Gerrard, B. (2007) Beyond Competitive Balance, in T. Slack and M.M. Parent (eds), International Perspectives on the Management of Sport, Burlington MA, USA, Elsevier. 149-172. La Liga. (2017). “Historical Stats, Standings”, available at: http://www.laliga.es/en/statistics- historical/standings ,(accessed November, 2017). Lee, Y.H. and Fort, R. (2012) Competitive Balance: Time Series Lessons from the English Premier League. Scottish Journal of Political Economy, 59(3), 266-282. Ligue 1. (2017). “League Table”, available at: http://www.ligue1.com/ligue1/classement ,(accessed November, 2017) Lindholm, J. (2010). The Problem with Salary Caps under European Union Law: The Case against Financial Fair Play. Texas Review of Entertainment and Sports Law, 12(2), 189-213. Lütkepohl, H. (2006). Vector Autoregressive Models. Palgrave Handbook in Econometrics, Vol. 1, Basingstoke, Hampshire and New York: Palgrave Macmillan, 477–510. Madden, P. (2012). Welfare Economics of Financial Fair Play in a Sports League with Benefactor Owners. Journal of Sports Economics, 16(2), 159-184. Manasis, V., Ntzoufras, I., and Reade, J. (2015). Measuring Competitive Balance and Uncertainty of Outcome Hypothesis in European Football. 1-24. Michie, J. and Oughton, C. (2004). Competitive balance in football: Trends and effects. The Sports Nexus Technical Report, 1-38. Morrow, S. (2003). The People’s Game? Football, Finance and Society, Palgrave Macmillan, New York, NY, United States of America. Muller, J., Lammert, J. and Hovemann, G. (2012). The Financial Fair Play Regulations of UEFA: An Adequate Concept to Ensure the Long-term Viability and Sustainability of European Club Football?. International Journal of Sport Finance, 7(2), 117-140.
26 Neale, W. (1964). The Peculiar Economics of Professional Sports. Quarterly Journal of Economics, 78(1), 1-14. Peeters, T., & Szymanski, S. (2012). Vertical restraints in soccer: Financial fair play and the English Premier League. University of Antwerp Working Paper no. 2012028. Peeters, T. and Szymanski, S. (2014). Financial Fair Play in European Football. Economic Policy, 29 (78), 343–390. Premier League. (2017). “Tables”, available at: https://www.premierleague.com/tables, (accessed November 2017). Preuss, H., Haugen, K., and Schubert, M. (2014). UEFA Financial Fair Play: The Curse of Regulation. European Journal of Sport Studies, 2(1), 33-51. Rottenberg, S. (1956). The Baseball players’ labour market. Journal of Political Economy, 64(3), 242-258. Sass, M. (2012). Long-term Competitive Balance under UEFA Financial Fair Play Regulations. Universitat Magdeburg Working Paper No. 5, 1-11. Sass, M. (2014). Glory Hunters, Sugar Daddies and Long-term Competitive Balance under UEFA Financial Fair Play. Journal of Sports Economics, 17(2), 148-158. Serby, T. (2014). UEFA’s Financial Fair Play Regulations: the devil is in the detail. 6-11. Serie A. (2017). “League Table”, available at: http://www.legaseriea.it/en/serie-a- tim/league-table, (accessed November, 2017). Stock, J. H. and Watson, M. W. (2012). Introduction to econometrics. Boston: Pearson/Addison Wesley. Storm, R.K. and Nielsen, K. (2012). Soft Budget Constraints in Professional Football. European Sport Management Quarterly, 12(2), 183-201. Szymanski, S. (2014). Fair is Foul: A Critical Analysis of UEFA Financial Fair Play. International Journal of Sport Finance, 9(3), 218-229. Union of European Football Associations. (2011). The European Club Licensing Benchmarking Report Financial Year 2011, UEFA, Nyon. Union of European Football Associations. (2015a). UEFA Financial Fair Play; All you need to know. UEFA, Nyon.
27 Union of European Football Associations. (2015b). UEFA Club Licensing and Financial Fair Play Regulations 2015. UEFA, Nyon. Vöpel, H. (2011). Do We Really Need Financial Fair Play in European Club Football? An Economic Analysis. CESifo DICE Report, 9(3), 54–59. Vöpel, H. (2013). Is Financial Fair Play Really Justified? An Economic and Legal Assessment of UEFA’s Financial Fair Play Rules. HWWI Policy Paper, 1-30. Quirk, J. and Fort, R. D. (1992) Pay Dirt: The Business of Professional Team Sports. Princeton New Jersey: Princeton University Press. Zimbalist, A.S. (2002). Competitive Balance in Sports Leagues: An Introduction. Journal of Sports Economics, 3(2), 111-121.
28 Appendix Data for the C5I from 1992/93-2016/17
C5I Season Premier League La Liga Bundesliga Ligue 1 Serie A Average
1992/93 127,5159236 142,1103582 131,622276 136,7588933 129,5823096 133,5179521 1993/94 139,3569132 134,2359768 122,166065 135,4330709 138,8697789 131,644163 1994/95 140,9265178 130,6042885 141,0576923 130,3588749 136,1236623 133,1320883 1995/96 137,0441459 138,7261146 128,4444444 131,1219512 134,8926014 132,1402071 1996/97 133,5952849 141,3417522 134,5754717 130,859375 130,5882353 132,9761516 1997/98 130,9090909 126,2135922 126,6266507 131,843044 139,9277978 130,4254017 1998/99 138,9268293 131,1601151 137,3285199 142,0024125 130,7526882 133,7519318 1999/00 139,3129771 125,2918288 133,7515078 121,1455847 139,8058252 129,502134 2000/01 135,8999038 134,4860711 125,5123675 129,3413174 140,5301205 131,47716 2001/02 146,6794995 128,2001925 139,3411765 131,1057692 138,1949458 132,8716015 2002/03 137,5238095 134,1062802 129,274673 126,8858801 136,8523002 130,9270114 2003/04 139,5348837 132,4401914 139,6441281 137,5600384 147,3913043 136,9103172 2004/05 145,2427184 135,3846154 132,9425557 126,5873016 141,8719212 132,8604635 2005/06 146,0018815 137,9710145 142,7737226 131,1154599 147,7707006 137,4293642 2006/07 141,2667946 134,7408829 135,1609058 123,828125 142,9149798 132,8321634 2007/08 152,3076923 140,9674235 134,5714286 130,46875 143,5797665 135,4206584 2008/09 151,1025887 138,8836329 138,1990521 138,5214008 140,861244 136,7962507 2009/10 148,0193237 149,6650718 135 138,063279 140,2697495 138,1028048 2010/11 136,8318756 142,8840716 137,2631579 129,1089109 138,4615385 135,0467205 Start FFP 2011/12 145,9407832 139,9617591 143,1377246 139,5348837 137,2434018 137,4787385 2012/13 150 148,4848485 140,5714286 135,1456311 144,2940039 139,2023671 2013/14 151,7890772 150,6641366 145,8548009 143,4108527 148,5714286 148,0580592 2014/15 143,2664756 159,0085796 139,5215311 139,5437262 139,3879566 144,1456538 2015/16 137,0764763 150,7633588 141,5348288 135,2713178 146,9856459 142,3263255 2016/17 154,5454545 152,2359657 138,1990521 148,7571702 153,2075472 149,389038
29 Data for HICB from 1992/93-2016/17
HICB Season Premier League La Liga Bundesliga Ligue 1 Serie A Average
1992/93 103,151497 109,034954 105,982916 107,0865035 106,1425061 106,2796753 1993/94 107,1300441 105,900355 104,9016669 108,1863414 108,7178311 106,9672477 1994/95 107,8537599 106,8742899 109,85288 105,7517632 110,0159625 108,0697311 1995/96 108,045579 106,6016877 105,0096022 105,0022606 108,5981511 106,6514561 1996/97 105,3184139 107,0213525 106,6477172 106,8992615 106,5203287 106,4814148 1997/98 105,3767084 106,0608917 104,2169569 106,2596054 111,0166951 106,5861715 1998/99 107,0886377 106,4651436 108,5195949 108,1009428 106,4194962 107,318763 1999/00 108,9279471 103,691956 107,6294924 102,856557 110,0769394 106,6365784 2000/01 106,8714736 105,47938 104,4612868 104,6270573 109,034693 106,0947781 2001/02 109,8024031 103,9775975 109,6143945 104,8498752 109,8906541 107,6269849 2002/03 108,092517 105,7686294 105,3179712 105,1595838 109,0379846 106,6753372 2003/04 108,2191575 105,438978 108,6890997 106,8635512 113,3322831 108,5086139 2004/05 110,4533886 107,2448225 108,1396739 103,9658919 107,1397025 107,3886959 2005/06 111,6473107 107,6356508 109,1427354 106,7436169 111,6971164 109,373286 2006/07 108,7934395 106,3472357 106,4906431 104,0306091 110,3935485 107,2110952 2007/08 113,683432 112,1394234 107,4030612 105,9455872 108,7980136 109,5939035 2008/09 111,6165938 107,16859 109,1316457 107,8138957 108,4279206 108,8317292 2009/10 112,6355341 112,0358966 107,8974436 107,9653518 107,6584955 109,6385443 2010/11 105,8610311 109,4924941 106,5952601 105,1387119 107,8883136 106,9951622 Start FFP 2011/12 110,5427888 109,7360792 110,213776 107,4680007 107,9626078 109,1846505 2012/13 111,3289466 110,7344754 111,1530612 106,4756339 110,4011014 110,0186437 2013/14 112,5155607 111,4395688 112,6524031 109,2745328 113,0376417 111,7839414 2014/15 109,2656601 114,9599101 108,0830338 107,5843225 109,5940337 109,897392 2015/16 108,4876707 111,3389371 110,4549845 108,1553092 110,3509535 109,757571 2016/17 113,4283173 114,6350583 108,3483075 110,6756505 114,2969028 112,2768473
30 Summary statistics: Premier League:
Summarize PL Variable Obs Mean St. Dev. Min Max
C5I 25 142,0247 6,999138 127,5159 154,5455
HICB 25 109,0455 2,722929 102,1515 113,6834
La Liga:
Summarize LaLiga Variable Obs Mean St. Dev. Min Max
C5I 25 139,2213 8,731857 125,2918 159,0086 HICB 25 108,2889 3,109332 103,692 114,9599 Bundesliga:
Summarize BL Variable Obs Mean St. Dev. Min Max
C5I 25 135,763 5,962657 122,1661 145,8548
HICB 25 107,862 2,186882 104,217 112,6524
Ligue 1:
Summarize L1 Variable Obs Mean St. Dev. Min Max
C5I 25 133,7509 6,382027 121,1456 148,7572 HICB 25 106,5152 1,811999 102,8566 110,6757 Serie A:
Summarize SA Variable Obs Mean St. Dev. Min Max
C5I 25 140,3573 5,778707 129,5823 153,2075 HICB 25 109,458 2,115926 106,1425 114,2969
31 Average: Average (before start FFP):
Summarize Avg Variable Obs Mean St. Dev. Min Max
(before FFP) C5I 19 133,5666 2,4527 129,5021 138,1028
HICB 19 107,5226 1,148387 106,0948 109,6385
Augmented Dicky-Fuller Tests:
Augmented Dicky-Fuller Test Measure Test Statistic 1% Critical Value 5% Critical Value
C5I (with trend) -5.308 (0.0001) -4.380 -3.600 HICB (with trend) -4.791 (0.0005) -4.380 -3.600 AR(p) models: C5I
Summarize Avg Variable Obs Mean St. Dev. Min Max
C5I 25 135,9346 5,260572 129,5021 149,389
HICB 25 108,2339 1,725759 106,0948 112,2768
C5I Coef. Std. Err. t-value P>t 90% LB 90% UB
L1. 0,117192 0,2045518 0,57 0,576 -0,2430872 0,4774713 L2. 0,6932336 0,2305942 3,01 0,009 0,2870857 1,099381 const. 25,69195 37,299130 0,69 0,502 -40,00339 91,38728 Number of obs 17 AIC 4,453847 HQIC 4,468462 SBIC 4,600884 R-squared 0,4240 Prob>F 0,0210 RMSE 2,07216
32
C5I Coef. Std. Err. z-value P>z 90% LB 90% UB
L1. -0,0279966 0,2537571 -0,11 0,912 -0,5253515 0,4693582 L2. 0,7280976 0,2148221 3,39 0,001 0,307054 1,149141 L3. 0,2417745 0,302029 0,8 0,423 -0,3501915 0,833741 const. 8,32599 41,184760 0,2 0,84 -72,39465 89,04663 Number of obs 16 AIC 4,616753 HQIC 4,626644 SBIC 4,8099 R-squared 0,4475 Prob>chi2 0,0047 RMSE 2,18864
C5I Coef. Std. Err. z-value P>z 90% LB 90% UB
L1. -0,0203089 0,2684459 -0,08 0,94 -0,5464533 0,5058354 L2. 0,7548998 0,2731203 2,76 0,006 0,2185939 1,2892060 L3. 0,2295596 0,3259942 0,7 0,481 -0,4093773 0,8684964 L4. -0,0742637 0,3195244 -0,23 0,816 -0,7005201 0,551993 const. 15,38685 49,764300 0,31 0,757 -82,14940 112,92310 Number of obs 15 AIC 4,837619 HQIC 4,835105 SBIC 5,073635 R-squared 0,4389 Prob>chi2 0,0192 RMSE 2,38517
33
AR(p) models: HICB
HICB Coef. Std. Err. t-value P>t 90% LB 90% UB
L1. 0,0106542 0,2278496 0,05 0,963 -0,3906597 0,411968 L2. 0,6265469 0,2447437 2,56 0,023 0,1954773 1,057616 const. 39,1723 29,74425 1,32 0,209 -13,21655 91,56115 Number of obs 17 AIC 3,011779 HQIC 3,026395 SBIC 3,158816 R-squared 0,3449 Prob>F 0,0518 RMSE 1,00759
HICB Coef. Std. Err. z-value P>z 90% LB 90% UB
L1. 0,1042406 0,2673334 0,39 0,697 -0,4197233 0,6282045 L2. 0,7027317 0,2212451 3,18 0,001 0,2690994 1,136364 L3. -0,1247181 0,2941869 -0,42 0,672 -0,7013138 0,451878 const. 34,22049 29,793840 1,15 0,251 -24,17437 92,61535 Number of obs 16 AIC 3,090886 HQIC 3,100777 SBIC 3,284033 R-squared 0,4184 Prob>chi2 0,0093 RMSE 1,02055
34
HICB Coef. Std. Err. z-value P>z 90% LB 90% UB
L1. 0,140309 0,2597627 0,54 0,589 -0,3688166 0,6494346 L2. 0,3806778 0,2361124 1,61 0,107 -0,082094 0,8434496 L3. -0,1602284 0,2939129 -0,55 0,586 -0,736287 0,4158302 L4. 0,6163507 0,2866539 2,15 0,032 0,0545195 1,178182 const. 2,76731 29,793840 0,09 0,932 -60,84957 66,38418 Number of obs 15 AIC 2,988796 HQIC 2,986282 SBIC 3,224813 R-squared 0,5637 Prob>chi2 0,0007 RMSE 0,94635
